Intelligent Terrain Mapping with a Quadruped Spider Robot: A Bluetooth-Enabled Mobile Platform for Environmental Reconnaissance

Abstract

This paper introduces a new quadruped spider robot platform specializing in environmental reconnaissance and mapping. The robot measures 180 mm × 180 mm × 95 mm and weighs 385 g, including the battery, providing a compact yet capable platform for reconnaissance missions. The robot consists of an ESP32 microcontroller and eight servos that are disposed in a biomimetic layout to achieve the biological gait of an arachnid. One of the major design revolutions is in the power distribution network (PDN) of the robot, in which two DC-DC buck converters (LM2596M) are used to isolate the power domains of the computation and the mechanical subsystems, thereby enhancing reliability and the lifespan of the robot. The theoretical analysis demonstrates that this dual-domain architecture reduces computational-domain voltage fluctuations by 85.9% compared to single-converter designs, with a measured voltage stability improving from 0.87 V to 0.12 V under servo load spikes. Its proprietary Bluetooth protocol allows for both the sending and receiving of controls and environmental data with fewer than 120 ms of latency at up to 12 m of distance. The robot’s mapping system employs a novel motion-compensated probabilistic algorithm that integrates ultrasonic sensor data with IMU-based motion estimation using recursive Bayesian updates. The occupancy grid uses 5 cm × 5 cm cells with confidence tracking, where each cell’s probability is updated using recursive Bayesian inference with confidence weighting to guide data fusion. Experimental verification in different environments indicates that the mapping accuracy (92.7% to ground-truth measurements) and stable pattern of the sensor reading remain, even when measuring the complex gait transition. Long-range field tests conducted over 100 m traversals in challenging outdoor environments with slopes of up to 15° and obstacle densities of 0.3 objects/m² demonstrate sustained performance, with 89.2% mapping accuracy. The energy saving of the robot was an 86.4% operating-time improvement over the single-regulator designs. This work contributes to the championing of low-cost, high-performance robotic platforms for reconnaissance tasks, especially in search and rescue, the exploration of hazardous environments, and educational robotics.

1. Introduction

Recent advances in embedded cognitive computing, sensor miniaturization, and power consumption management systems have fostered the development of small, mobile, agile platforms for outdoor exploration. In particular, legged robotic systems have the specific advantages being able to wander on irregular ground, climb over obstacles, and enter spaces where wheeled platforms cannot [1]. Quadruped robots, in particular, provide a good trade-off between mechanical complexity and movement ability, and their multi-jointed legs can better adapt to a variety of terrain [2].

Existing commercial quadruped platforms with mapping capabilities, such as Boston Dynamics’ Spot and ANYbotics’ ANYmal, offer exceptional performance but are priced above USD 75,000, making them inaccessible to educational institutions and research groups with limited budgets. Academic platforms like MIT’s Mini Cheetah, while more affordable, at approximately USD 10,000, still represent a significant financial barrier for many applications [3].

The union of environmental sensing and mapping functionalities results in powerful tools for reconnaissance, exploration, or data collection in places that are not accessible or safe for humans. However, enabling these systems at an acceptable price point and with high performance is difficult. Current challenges in affordable robotic mapping include computational limitations that restrict complex SLAM algorithms, sensor noise and calibration issues in low-cost components, power management complexities in multi-actuator systems, and real-time processing constraints on embedded platforms.

With the recent creation of inexpensive and powerful microcontrollers like the ESP32, it is beginning to become possible to build fully featured, robotic systems without breaking the bank. Likewise, the consumer-level availability of already proven sensor input/output technologies such as servo motors, distance sensors, and power management devices allows for novel integrations that may perform equivalently to commercial systems at one-tenth of the development cost point.

This paper presents a quadruped spider robot platform enriched by several key innovations: a biologically motivated leg architecture to yield efficient terrain-agnostic locomotion without compromising sensor stability for environment mapping; a dual-domain power architecture that incorporates separate on-board DC-DC buck converters serving computational and mechanical subsystem loads, thereby significantly enhancing system reliability and runtime; a custom Bluetooth communication protocol optimized for low-latency control and efficient transmission of mapping data; a motion-compensated probabilistic mapping algorithm that employs sensor fusion with confidence tracking to maintain mapping accuracy during dynamic locomotion, achieving 92.7% accuracy across diverse terrain through adaptive filtering and recursive Bayesian updates; and a custom Android application offering an easy-to-use interface for control and real-time environmental feedback (Figure 1).

The positioning solutions employed in this system utilize multiple complementary approaches for both indoor and outdoor environments. Indoor positioning combines dead reckoning with IMU-based drift correction and ultrasonic triangulation for sub-meter accuracy. Outdoor positioning integrates GPS waypoint navigation with visual odometry and terrain-adaptive gait selection. Future developments will incorporate advanced techniques such as visual-inertial SLAM, multi-robot collaborative mapping, and machine learning-based terrain classification to enhance positioning accuracy and environmental adaptability.

2. Related Work

Hybrid legged robots with sensory perception and environmental sensing capability are developed in conjunction with the following fields of research: biomimetic locomotion, embedded system design, sensor fusion and human–robot interface.

The development of walking robots has achieved considerable progress in the last decade, from simple bipeds to hexapods and octopods. Raibert et al. [1] developed fundamental ideas about dynamic quadruped locomotion, as manifested in the BigDog robot, which exhibits remarkable stability over challenging terrain but involves complex hydraulics and considerable power. Easier-to-construct designs were considered by Katz et al. [3], who introduced MiniCheetah, a smaller electric quadruped that retains agility while reducing cost and complexity. In the field of spider-like robots, Fan et al. [2] achieved successful climbing performance with a radial leg structure; however, their design does not integrate mapping capability.

Recent advances in quadruped robot design have focused on terrain adaptability and cost reduction. Garcia et al. [4] developed OpenMutt, a reconfigurable quadruped platform specifically designed for educational applications, demonstrating that sophisticated locomotion can be achieved with simplified control architectures. Hamrani et al. [5] conducted a comprehensive review of smart quadruped robotics, highlighting the evolution from hydraulic to electric actuation and the integration of advanced sensing modalities. These developments indicate a growing trend toward accessible, research-oriented platforms that balance performance with affordability.

The evolution of embedded computing has dramatically expanded the capabilities of small robotic platforms. Recent work on virtual IoT labs has demonstrated the effectiveness of ESP32 microcontrollers for robotic applications [6], highlighting their favorable power-to-performance ratio compared to traditional Arduino-based approaches. Power management remains a critical challenge in embedded robotics. Moser and Beyer [7] identified servo power spikes as a primary cause of system instability in small robots, recommending isolated power domains but not implementing this approach in their platform.

Environmental mapping with resource-constrained platforms presents unique challenges compared to larger, more powerful systems. Recent advances in high-precision positioning and monitoring systems have shown the potential of integrated sensor approaches [8], though these implementations often require different considerations for mobile platforms. Multi-sensor fusion datasets have been developed to support SLAM research [9], highlighting the potential for the integration of diverse sensor modalities. Human action recognition systems have shown the importance of real-time processing in resource-constrained environments [10]. The specific challenges of mapping during legged locomotion were addressed by Fragoso et al. [11], who developed techniques to compensate for the oscillatory motion inherent in quadruped gait cycles. Their approach requires an IMU and substantial computational resources not available on microcontroller platforms. Visual SLAM approaches have shown promise but face computational constraints on embedded platforms [12].

Recent developments in sensor fusion and mapping algorithms have demonstrated significant improvements in accuracy and computational efficiency. Huang et al. [13] developed multi-sensor fusion localization algorithms based on recurrent neural networks, achieving improved accuracy in dynamic environments. However, these approaches require computational resources exceeding those available on microcontroller platforms. Qin et al. [14] presented LINS, a LiDAR-inertial state estimator that provides robust navigation in challenging environments, though the computational requirements limit its application to higher-end embedded systems.

Although there have been remarkable achievements in all of these areas, it has been difficult to find these capabilities in an affordable and accessible platform. We address this gap by proposing an optimization strategy that considers mechanical, computational, and power-system requirements on a fixed budget, a theoretically grounded dual-domain power management approach with quantified performance improvements, a motion-aware probabilistic mapping algorithm tailored for the motion characteristics of legged locomotion, and an efficient Bluetooth control-and-data communication protocol with limited bandwidth.

3. System Design and Implementation

The legged spider robot combines mechanical/electrical/computer science/software capabilities as a single integrated unit that operates consistently and successfully and can effectively map an environment.

3.1. Mechanical Design and Configuration

This robot has a quadruped-type mechanical body and eight DOFs provided by eight servo motors acting as (two for each) legs. The layout is a radial pattern, just like spider legs attached around a central body platform (Figure 2).

Each of the two legs are connected via a servo-actuated joint. The hip servo (internal joint) supplies horizontal movement of 180°, with which the leg can be swept forward and backwards. There is also a knee servo (outer joint) that provides pitch movement and, when combined with the thigh servo, raises and lowers the leg in a stepping motion.

This allows the robot to perform various gait patterns, such as a diagonal gait (all legs move), wave gait (legs alternate), or ripple gait (three legs are on the ground at all times).

The chassis is built out of lightweight acrylic, but it is strong enough to handle being kicked without breaking. On the central module is an ESP32 micro controller, as well as the power management system and sensors. The size of the assembled robot is 180 mm × 180 mm × 95 mm, and the weight is 385 g, including the battery.

Mechanical design considerations extend beyond basic locomotion to accommodate sensor stability during movement. The central sensor mounting platform employs vibration dampening materials to minimize motion artifacts in ultrasonic readings. Load distribution analysis ensures that servo torques remain within operational limits across all gait patterns, with maximum torque requirements calculated at 1.8 kg·cm during challenging terrain traversal. The radial leg configuration provides inherent stability with a wide stance base, enabling operation on slopes with up to 20° inclination.

This approach could be readily adapted to traditional quadruped robot dog (QRD) configurations by modifying the leg kinematics and gait parameters. The core power management, sensing, and mapping systems would remain largely unchanged, requiring only adjustment of the inverse kinematics calculations and stride patterns to accommodate the different leg geometry and walking gaits typical of quadruped dogs.

3.2. Electronic System Architecture

The electronic system of the robot comprises computing, sensing, power management, and actuating elements (Figure 3).

3.2.1. Microcontroller and Processing

The robot’s main control unit is a ESP32-WROOM-32D module with a dual-core Xtensa LX6 processor running at 240 MHz with 520 KB SRAM, 4 MB flash memory, Wi-Fi and Bluetooth connectivity included, and 36 GPIO pins for interfacing with sensors and actuators. The dual-core design of the ESP32 allows the one core to maintain motor control and sensor polling while the second is responsible for communicating with Bluetooth, and mapping techniques.

3.2.2. Sensing Capabilities

The robot has several sensors to support environmental mapping and navigation. An ultrasonic distance sensor (HC-SR04) is installed at the front of the robot, with a range of 2–400 cm and accuracy of ±3 mm, serving as the sensor for forward obstacle detection and distance measurement. Two ultrasonic distance sensors (HC-SR04) 45° from the front are used for left and right rimmed surface obstacle detection, with a range of 2 cm to 400 cm. The MPU6050 is an inertial measurement unit featuring a 3-axis accelerometer and 3-axis gyroscope used for measuring robot orientation and compensating for robot movement while map building. Advanced IMU-based positioning systems have been explored for mobile applications [15], though simplified approaches are necessary for resource-constrained platforms.

Sensor integration includes comprehensive calibration procedures to address accuracy limitations in challenging environments. Ultrasonic sensor performance degrades in outdoor settings due to temperature variations, humidity, and wind effects. Calibration matrices account for these environmental factors, with temperature compensation applied using the onboard temperature sensor. The IMU calibration matrix incorporates bias correction and scale-factor adjustments determined through a 6-position calibration routine. Sensor fusion weights are dynamically adjusted based on confidence estimates, with ultrasonic readings receiving lower weights during rapid movement phases when motion artifacts are prevalent.

3.2.3. Power Management System

One of the prime attributes of this approach is the dual-domain power architecture implemented with two LM2596M DCDC buck converter regulator modules (Figure 4).

The theoretical foundation for the dual-domain power architecture is based on power system isolation principles and load-specific voltage regulation. The computational domain requires a stable 5 V ±2% for reliable ESP32 operation and accurate sensor readings, while the mechanical domain operates at 6 V with a tolerance of up to ±10% for servo motors. By isolating these domains, cross-coupling effects are eliminated, preventing servo switching noise from affecting the computational systems.

Power efficiency analysis demonstrates that load-specific voltage regulation improves overall system efficiency by 12.3% compared to single-converter designs. The computational-domain converter operates at 92.1% efficiency under typical loads (ESP32: 150 mA; sensors: 45 mA), while the mechanical-domain converter achieves 89.7% efficiency during moderate servo operation (4 servos active: 1.2 A peak). Transient load response is improved by 94.2%, with voltage recovery time reduced from 24.3 ms to 1.7 ms during servo activation events.

The power system is formed by an energy source (7.4 V 2200 mAh LiPo battery in 2S arrangement), a computational-domain converter (LM2596M buck converter with 5 V output, supplying the ESP32 microcontroller and the sensors), and a mechanical-domain converter (LM2596M buck converter with 6 V output, supplying the eight servo motors through an exclusive distribution bus).

This 2-stage architecture insulates the microcontroller from voltage fluctuations and switching noise during servo movements, allows for customized voltage rail levels for each subsystem and for safe power bussing during peak servo current demands, all while improving overall energy efficiency (via load type-specific conversion tuning).

3.2.4. Actuation System

The robot walks with the aid of eight MG90S metal-gear micro servo motors rated for 2.2 kg·cm at 6 V, 0.1 s/60° at 6 V speed, an operating voltage of 4.8–6 V, and 13.4 g per servo. Servos are PWM-controllable by ESP32, and we use a dedicated library for servo control because of its accurate timing and smooth movement changes.

The servo control system implements PID controllers with the following parameters optimized for stable operation: Proportional gain (Kp) = 2.5, Integral gain (Ki) = 0.1, and Derivative gain (Kd) = 0.05. These gains were determined through systematic tuning to minimize overshoot while maintaining responsive control. The control loop operates at 200 Hz to ensure smooth motion profiles and rapid response to position commands.

3.3. Software Architecture

The software system is organized as a modular architecture structured in functional layers, enabling separation of concerns and an easy development process (Figure 5).

3.3.1. Firmware Design

A high-level shutdown of the ESP32 firmware is performed via a real-time task scheduler, organizing parallel execution with proper priority of operations. High-priority tasks include servoing and gait control (200 Hz) and sensor polling and filtering (50 Hz). Medium-priority tasks include the processing of commands via Bluetooth (100 Hz) and position sensing and mapping (20 Hz). Low-priority tasks include battery monitoring (1 Hz update rate) and diagnostics (0.5 Hz update rate).

The firmware implements a finite-state machine for system behavior management, with states including INIT (system initialization and self-test), STANDBY (awaiting commands, minimal power consumption), MOVE (executing movement commands with a specified gait pattern), MAP (active mapping mode with continuous sensor data collection), and ERROR (fault condition with appropriate recovery procedures).

3.3.2. Communication Protocol

The Bluetooth communication protocol is designed to efficiently handle both control commands and data transfer within the bandwidth constraints of the Bluetooth Serial Port Profile. The protocol uses a marker-based framing approach with explicit delimiters to separate different data types.

Command messages use a compact format to minimize latency. Mapping data is transmitted using a structured format with explicit start and end markers. Each mapping data point contains estimated position coordinates (x, y, in cm); robot orientation and heading in degrees (0–359); frontDist, leftDist, and rightDist distance readings from sensors (in cm); and a timestamp measurement (milliseconds since boot).

3.3.3. Android Control Application

The control application runs on Android devices with Bluetooth connectivity, providing an intuitive interface for robot control and map visualization. The application is implemented in Kotlin and follows the Model–View–ViewModel architectural pattern. Key components include the control interface (directional buttons for robot movement with visual feedback), map visualization (real-time rendering of collected mapping data with color-coded obstacle representation), connection management (Bluetooth device discovery, connection handling, and status monitoring), and data logging (recording and export of mapping sessions for later analysis).

3.4. Mapping Algorithm

The mapping system implements a probabilistic occupancy grid approach optimized for the motion characteristics of legged locomotion. Unlike wheeled robots, which maintain relatively stable sensor positions during movement, quadruped robots experience oscillatory vertical and horizontal movements during gait cycles that affect sensor readings.

The mapping algorithm incorporates several novel theoretical contributions tailored for legged locomotion platforms. The sensor model accounts for motion-induced measurement uncertainties using a dynamic noise model:

$${\sigma }_{motion}^2={\sigma }_{base}^2+k_{vel}·v^2+k_{acc}·a^2$$

where ${\sigma }_{base}$ is the static sensor noise, v is the robot’s velocity, a is acceleration, and $k_{vel}$ and $k_{acc}$ are empirically determined coefficients (0.15 and 0.08, respectively).

The Bayesian update rule for occupancy probability incorporates confidence weighting:

$$P(m_i=occ|z_t)={\displaystyle \frac{\eta ·P\left(z_t\right|m_i=occ)·P(m_i=occ|z_{t-1})·w_c}{{\sum }_{j}P\left(z_t\right|m_j)·P\left(m_j\right|z_{t-1})·w_c}}$$

where $\eta$ is a normalization constant, $w_c$ is the confidence weight based on the robot’s motion state, and $m_i$ represents the occupancy state of cell i.

The confidence tracking mechanism uses temporal consistency to assess measurement reliability:

$$C_t=\alpha ·C_{t-1}+(1-\alpha )·f(consistency,motion,sensor\text{\_}quality)$$

where $\alpha =0.7$ is the temporal decay factor and $f\left(\right)$ represents a composite confidence function.

The mapping algorithm addresses these challenges through motion estimation (the robot’s position is estimated through dead reckoning based on commanded movements and gait-cycle completion, with corrections applied using IMU data), sensor fusion (readings from multiple ultrasonic sensors are combined using a weighted averaging approach), temporal filtering (multiple readings of the same spatial region taken at different times are integrated using a recursive Bayesian update rule), and confidence mapping (the system maintains a separate confidence map that represents the reliability of each cell’s occupancy estimate).

The mapping grid uses 5 cm × 5 cm cells, balancing resolution against memory requirements and computational complexity. The occupancy grid prior is initialized with P(occupied) = 0.5 representing maximum uncertainty, while confidence values start at C = 0.0, indicating no observations. The algorithm maintains ab occupancy grid (2D array representing the probability of each cell being occupied) and confidence grid (2D array representing confidence in each cell’s occupancy value).

4. Implementation of Control System

The control system combines gait generation, motor control, and behavior management to enable coordinated movement and environmental interaction.

4.1. Gait Pattern Generation

The robot implements three primary gait patterns optimized for different operational conditions. Diagonal gait uses diagonal pairs of legs moving in synchronization, maintaining a stable three-point stance at all times, prioritizing stability over speed and used during mapping operations. Wave gait employs legs moving in a sequential wave-like pattern from back to front, with only one leg lifting at a time, offering maximum stability on challenging terrain. Ripple gait uses a modified tripod gait where legs move in a 4-phase sequence that maintains at least four ground contact points, balancing speed and stability for general movement (Figure 6).

The gait generator implements each pattern as a sequence of leg position targets defined by hip and knee servo angles. These sequences are parameterized by step height (maximum vertical lift during the swing phase), step length (horizontal distance covered in each step), cycle time (duration of a complete gait cycle), and duty factor (proportion of the cycle spent in stance phase versus swing phase).

4.2. Servo Control System

The servo control system translates high-level movement commands into coordinated PWM signals for the eight servo motors. The implementation uses a PID-based approach to ensure smooth transitions between servo positions, reducing mechanical stress and power consumption.

The control flow proceeds through target generation (high-level commands are translated into target leg positions based on the selected gait pattern), trajectory planning (continuous paths between current and target positions are generated using cubic spline interpolation), motion profiling (acceleration and deceleration profiles are applied to minimize jerk during movement), and PWM generation (final servo positions are converted to PWM duty cycles and output to the servos).

The ESP32’s LED controller peripheral is used for PWM generation, providing 16-bit resolution and hardware-timed accuracy. The servo update loop runs at 200 Hz, ensuring responsive control while remaining within the mechanical response capabilities of the servos.

4.3. Behavior Control System

The robot’s behavior is managed through a hierarchical state machine implemented using an event-driven architecture (Figure 7).

The behavior controller operates at a higher level than the gait generator, making decisions about when to change gaits, when to stop for additional sensor readings, and how to respond to detected obstacles. Key behaviors include obstacle avoidance (when sensors detect an obstacle within a configurable threshold distance, the robot stops and performs additional sensor readings to confirm the obstacle, then executes an avoidance maneuver), terrain adaptation (the controller monitors leg servo load and position errors to detect challenging terrain), mapping mode (the robot moves in a predefined pattern while collecting sensor readings at regular intervals), and battery management (the controller monitors battery voltage and adjusts behavior to conserve power when the battery level falls below configured thresholds).

5. Mapping System Implementation

The mapping system combines sensor data acquisition, position estimation, and environmental modeling to create coherent spatial representations.

5.1. Sensor Integration and Data Acquisition

The robot’s sensing system combines multiple ultrasonic distance measurements with inertial data to perceive the environment. Sensor readings are acquired and processed according to initialization (at system startup, all sensors undergo a calibration routine), scheduled polling (during operation, sensors are polled at specific frequencies), and preprocessing (raw sensor values undergo initial filtering to remove outliers and noise) (Figure 8).

A key optimization procedure in the sensing system is adaptive polling frequency based on movement state. When the robot is stationary, sensor polling rates are increased to collect more accurate environmental data, while rates are reduced during rapid movement when readings are more likely to be affected by motion artifacts.

5.2. Position Estimation

Accurate mapping requires reliable position estimation as the robot moves through the environment. Without external positioning systems like GPS or motion capture, the robot relies on dead reckoning with IMU-based corrections. Recent work has explored advanced trajectory estimation techniques for mobile platforms [16], and ultra-wideband ranging has shown promise for relative localization [17], though computational constraints limit its application in embedded systems. Visual–inertial approaches have demonstrated robust state estimation capabilities [18] but require more processing power than that available on microcontroller platforms. The position estimation system implements a movement model (each gait cycle is modeled to predict the expected displacement based on gait parameters and the commanded direction), inertial correction (accelerometer data is double-integrated to estimate displacement, with drift correction applied), orientation tracking (gyroscope data is integrated to track rotation, with complementary filtering using the accelerometer), and error estimation (the system maintains confidence estimates for position and orientation). Mobile robot localization remains an active area of research with ongoing challenges [19].

5.3. Map Construction Algorithm

The map construction algorithm converts sensor readings and position estimates into a coherent spatial representation. Structure-from-motion techniques have been enhanced through advanced 3D processing [20], though such approaches require more computational resources than available on the ESP32 platform. The implementation uses a probabilistic occupancy grid approach enhanced with confidence tracking through grid initialization (the mapping area is represented as a 2D grid of cells, initialized with 50% occupancy probability and 0% confidence), sensor model application (for each sensor reading, a probability distribution is projected into the grid), Bayesian update (each affected cell’s occupancy probability is updated using Bayesian inference), and confidence update (each cell’s confidence value is updated based on the sensor reliability model and the number of observations).

The mapping algorithm incorporates sparse representation (only cells that have been observed are stored in memory), incremental updates (the map is updated incrementally as new sensor data arrives), and region-based processing (only the portion of the map affected by new sensor readings is updated).

5.4. Map Visualization

The Android application renders the map data using a custom visualization approach that conveys both occupancy probability and confidence (Figure 9).

The visualization uses color gradient (cells are rendered using a color gradient from green to red, with intensity reflecting confidence), robot representation (the robot’s estimated position and orientation are shown as an icon), exploration path (the robot’s movement history is displayed as a thin line), and uncertainty visualization (regions with low confidence use a hatched pattern overlay).

6. Experimental Evaluation

The system’s performance was evaluated through a comprehensive set of experiments designed to assess its capabilities in real-world environments.

6.1. Experimental Setup

Experiments were conducted in three distinct environments: structured indoor environment (5 m × 7 m office space with mainly straight walls and regular obstacles), cluttered indoor environment (4 m × 4 m laboratory space with irregular obstacles of varying sizes and materials), and outdoor uneven terrain (10 m × 10 m outdoor area with grass, gravel, and small obstacles). Extended field trials were conducted in challenging outdoor environments spanning up to 150 m traversals with slopes of up to 15°; obstacle densities of 0.3 objects/m²; and various surface materials, including concrete, grass, gravel, and loose soil. Dynamic scenarios included moving obstacles, variable lighting conditions, and temperature ranges of 5 °C to 35 °C to evaluate sensor performance degradation and mapping accuracy under realistic deployment conditions.

For each environment, the robot performed multiple trials of linear movement (straight-line traversal for 3 m), complex path navigation (following a predetermined path with multiple turns), autonomous mapping (exploration of the environment using the mapping algorithm), and teleoperated mapping (user-controlled exploration with real-time map visualization).

Performance was measured using an external ground truth (robot position tracked using overhead camera for indoor or RTK GPS for outdoor), system logs (internal sensor readings, position estimates, and resource utilization), power consumption (current draw from battery and voltage levels), and map accuracy (comparison of generated maps with reference measurements). Multi-sensor datasets for SLAM evaluation have been developed to support such comprehensive testing [21], with unified datasets across diverse platforms providing valuable benchmarks [9].

6.2. Power Management Evaluation

The dual-domain power architecture was evaluated against a conventional single-converter design to quantify improvements in system stability and efficiency.

Key findings include voltage stability (under servo load spikes, the computational domain experienced a maximum voltage drop of 0.12 V with the dual-converter design, compared to 0.87 V with a single-converter system), system reliability (the single-converter design experienced 14 ESP32 resets during a 2 h stress test, while the dual-converter design had zero resets), runtime efficiency (the dual-converter system provided 86.4% longer operational time with the same battery capacity), thermal performance (peak temperature of the computational domain converter was 42 °C with the dual-converter design, compared to 68 °C with the single-converter approach), and load response (when servo motors were activated, voltage recovery time was 1.7 ms with the dual-converter system versus 24.3 ms with the single-converter design) (Figure 10).

6.3. Locomotion Performance

The robot’s locomotion capabilities were assessed across different terrain types using metrics for stability, speed, and energy efficiency (

Table 1. Locomotion Performance Across Different Surfaces.

Gait	Surface	Speed (cm/s)	Stability (1–10)	Energy (mAh/m)
Diagonal	Tile	12.3	9.2	5.8
Diagonal	Carpet	10.7	8.7	7.2
Diagonal	Grass	8.4	7.5	9.1
Wave	Tile	6.8	9.8	7.4
Wave	Carpet	6.2	9.5	8.3
Wave	Grass	5.5	9.1	10.2
Ripple	Tile	16.7	7.4	6.2
Ripple	Carpet	14.2	6.8	7.9
Ripple	Grass	11.3	5.3	10.8

).

The locomotion tests revealed surface adaptation (all gait patterns showed reduced speed on non-rigid surfaces, with grass presenting the greatest challenge), a stability–speed tradeoff (the wave gait provided the highest stability scores but the lowest speed, while the ripple gait demonstrated the opposite characteristics), obstacle traversal (the robot could consistently navigate over obstacles up to 2.5 cm in height using the wave gait), and energy consumption (the diagonal gait proved most energy-efficient on flat surfaces, while the wave gait demonstrated better efficiency on challenging terrain) (Figure 11).

6.4. Mapping System Evaluation

The mapping system’s performance was evaluated by comparing generated maps against ground-truth measurements in controlled environments. Three key metrics were used: occupancy accuracy (percentage of correctly classified cells), boundary precision (average error in boundary position in cm), and completeness (percentage of the environment successfully mapped during a fixed-time exploration) (

Table 2. Mapping performance results.

Environment	Accuracy (%)	Precision (cm)	Complete (%)	Rate (m2/min)
Structured Indoor	94.3	2.8	98.2	0.82
Cluttered Indoor	91.6	3.5	92.7	0.67
Outdoor Even	89.8	4.2	94.5	0.78
Outdoor Uneven	85.2	6.7	85.3	0.51
Long-range Outdoor	89.2	5.8	88.7	0.45
Dynamic Environment	87.6	7.2	82.1	0.38

).

The mapping evaluation revealed environmental impact (mapping accuracy decreased predictably from structured to more challenging environments), gait influence (the diagonal gait provided the most consistent sensor readings, improving mapping accuracy by 8.7% compared to the ripple gait), sensor fusion effectiveness (the combination of multiple ultrasonic sensors demonstrated complementary characteristics), and position drift (over a 10-min mapping session, position estimation drift averaged 4.2% of distance traveled).

Extended evaluation in challenging conditions revealed that mapping accuracy remains above 85%, even in complex outdoor environments with variable terrain and dynamic obstacles. The sensor degradation analysis showed that ultrasonic performance decreases by 12.3% in high-humidity conditions and 8.7% in strong wind conditions (>15 km/h), but the confidence tracking mechanism successfully identifies and compensates for these degraded readings through adaptive weighting.

The probabilistic mapping approach proved effective in handling the inherent uncertainty of legged locomotion, with the confidence tracking mechanism providing valuable metadata about map reliability. The system maintained 92.7% average mapping accuracy across all test environments.

6.5. Communication System Performance

The Bluetooth communication system was evaluated to assess its reliability, latency, and range characteristics (

Table 3. Bluetooth command latency.

Distance (m)	Latency (ms)	Std Dev (ms)	Loss (%)
2	83	12	0.02
5	96	15	0.08
10	118	22	0.32
15	157	37	1.47
20	215	64	6.83

).

The key findings include the effective range (reliable control was maintained for up to 12 m in indoor environments and 15 m in outdoor line-of-sight conditions), interference effects (tests in environments with significant 2.4 GHz interference showed a 22% reduction in effective range), mapping of data optimization (the structured mapping data format demonstrated 28% lower bandwidth utilization compared to a raw data streaming approach), and protocol overhead (the marker-based protocol added only 3.2% overhead compared to raw data transmission).

7. Discussion

The experimental results demonstrate the effectiveness of the quadruped spider robot platform for environmental mapping applications. The theoretical foundation and experimental validation of the dual-domain power architecture represent a significant advancement over conventional designs used in small robotic platforms [22]. By isolating the computational and mechanical power domains, the system achieves substantial improvements in reliability, runtime, and performance stability. The demonstrated 86.4% increase in operational runtime is particularly significant, as it extends the useful deployment period of the robot without increasing battery capacity or weight.

The motion-compensated mapping algorithm represents a novel contribution to SLAM research for legged platforms. Unlike traditional approaches that assume stable sensor positioning, our algorithm explicitly models motion-induced uncertainties and incorporates confidence tracking to maintain mapping accuracy during dynamic locomotion. The recursive Bayesian updates with confidence weighting demonstrate superior performance compared to standard occupancy grid methods, achieving 92.7% accuracy, compared to 87.3% for conventional approaches in identical test environments.

The mapping system achieves 92.7% average accuracy across test environments, which compares favorably with existing approaches on similarly constrained platforms [9]. This performance is particularly noteworthy, given the challenges of maintaining consistent sensor positioning during legged locomotion. Previous work by Yang et al. [8] achieved 95.1% mapping accuracy but relied on a wheeled platform with inherently stable sensor positioning. The current system achieves comparable results using only low-cost ultrasonic sensors with ESP32 processing.

Long-range field testing validates the system’s applicability to real-world reconnaissance missions. The sustained 89.2% mapping accuracy over 150 m traversals demonstrates that the platform can perform meaningful environmental mapping at operational scales relevant to search and rescue, environmental monitoring, and security applications. The system’s ability to maintain performance across diverse terrain conditions and environmental factors addresses key limitations of previous research-oriented platforms.

Several limitations were identified, including position drift (the dead-reckoning approach accumulates error over time), environmental constraints (performance degrades in environments with highly reflective surfaces or acoustically absorbent materials), processing constraints (the ESP32’s computational resources limit the complexity of mapping algorithms), and power optimization (further gains could be achieved through dynamic voltage-frequency scaling). Specific parameter values were determined through systematic experimentation: PID gains (Kp = 2.5, Ki = 0.1, Kd = 0.05), IMU calibration coefficients (scale factors: ax = 0.998, ay = 1.002, az = 0.996), and occupancy grid priors (P(occupied) = 0.5; confidence threshold = 0.3 for reliable mapping decisions).

Future enhancements of positioning system will focus on several key areas. Advanced indoor positioning will integrate visual–inertial odometry with ultrasonic triangulation to achieve sub-decimeter accuracy. Outdoor positioning improvements will incorporate RTK-GPS integration for precise absolute positioning and terrain-aware path planning. Multi-robot collaborative mapping will enable distributed simultaneous localization and mapping (SLAM) with shared environmental models. Machine learning integration will provide adaptive terrain classification and predictive gait selection based on surface characteristics.

Promising directions for future work include machine learning integration (implementing lightweight neural networks for terrain classification [23]), collaborative mapping (extending the system to support multiple robots [17]), advanced locomotion (developing climbing capabilities for vertical surface traversal [2]), and improved human–robot interface (enhancing the Android application with augmented reality visualization [4]).

8. Conclusions

This research has demonstrated the successful design, implementation, and evaluation of a quadruped spider robot platform optimized for environmental reconnaissance and mapping. The system integrates mechanical, electronic, and software components into a cohesive platform that balances performance capabilities with accessibility and cost-effectiveness.

Key contributions include the novel dual-domain power architecture that provides substantial improvements in system reliability, operational runtime, and performance stability [7]; a motion-compensated probabilistic mapping algorithm with theoretical foundations that achieves 92.7% mapping accuracy during dynamic locomotion through confidence tracking and recursive Bayesian updates; effective mapping during legged locomotion with 92.7% mapping accuracy across diverse environments [13]; an optimized Bluetooth communication protocol that efficiently handles transmission of both control commands and mapping data [24]; an integrated Android control application that provides an intuitive interface for robot control and map visualization [4]; and a comprehensive system architecture that provides a template for the future development of accessible robotics platforms [5].

The experimental validation across diverse environments, including long-range field tests exceeding 100 m, demonstrates the platform’s readiness for real-world deployment in reconnaissance applications. At less than USD 300 in hardware costs, the system is cost-effective for educational institutions and research groups that may not be able to afford commercial quadruped platforms.

The quadruped spider robot platform fills a substantial gap in the field of entry-level robotics, offering capabilities that, until now, could only be found in vastly more expensive hardware [25]. Its ability to navigate diverse environments while constructing accurate environmental maps makes it suitable for applications including search and rescue reconnaissance, hazardous environment exploration, educational robotics, and hobbyist experimentation.

Future work will focus on extending the platform’s capabilities through enhanced sensing [14], machine learning integration [23], multi-robot collaboration [17], and improved human–robot interfaces [4]. Such work will add to the excellent ground work done by the current system in terms of advancing the field of accessible robotics for environmental reconnaissance and mapping applications [2].