A Survey on Mechanical Solutions for Hybrid Mobile Robots

: This paper presents a survey on mobile robots as systems that can move in di ﬀ erent environments with walking, ﬂying and swimming up to solutions that combine those capabilities. The peculiarities of these mobile robots are analyzed with signiﬁcant examples as references and a speciﬁc case study is presented as from the direct experiences of the authors for the robotic platform HeritageBot, in applications within the frame of Cultural Heritage. The hybrid design of mobile robots is explained as integration of di ﬀ erent technologies to achieve robotic systems with full mobility.


Introduction
Mobile robotics is one of the main research topics in robotics, addressing robot design, navigation, perception, mapping, localization, motion planning, and control. A mobile robot is defined by its capacity of moving from one location to another arbitrarily distant one, by using one or more modes of locomotion. These locomotory structures include an extremely wide range of mechanical, chemical, mechatronic, electro-magnetic, pneumatic and hydraulic architectures, whose goal is to move the robot through a pre-defined environment. The earliest mobile robots were developed in the late 1960s [1,2], when the wheeled robot Shakey was developed as a testbed for DARPA-funded AI research at Stanford Research Institute [3,4]. At the beginning, mobile robots were only seen as tools to demonstrate AI and navigation systems. However, many research groups quickly grew more and more interested in studying locomotion itself, designing and inventing robots with various modes of locomotion. In the 1970s, the spotlight was on legged robots. The first legged system was the Phony Pony [5], which was developed in 1968 as a two-degrees-of-freedom (DoF) quadruped robot able to walk in a straight line only. In the same year, the General Electric Quadruped was invented and built [6]. Biped robots appeared some years later, when, in 1970, Kato started the WABOT project at Waseda University, Tokyo [7]. In the meanwhile, the Vietnam War and the First Persian Gulf War pushed military research groups to focus on Unmanned Aerial Vehicles (UAV) [8]. Several Autonomous Underwater Vehicles (AUV) were developed in the 1980s and 1990s, as reported in [9,10]. In the last decades, more and more modes of locomotion were considered for robots, and each of them was developed independently from the others. The designs of two different mobile robots often have nothing in common, and motion parameters and requirements may vary from task to task. Because of these issues, it is difficult to find a general classification of mobile robots and autonomous systems, and very few researchers have approached mobile robots with a wide perspective. Furthermore, some tasks could be performed only by combining two or more modes of locomotion, or they could require moving through different environments. In these cases, a mobile robot should be able to reconfigure itself for the new scenario in order to achieve its goals. This kind of robots is the focus of this article, which addresses the different fields, [45][46][47][48][49], has been growing and growing, and they will likely have a huge impact on everyday life in the next decade, [50][51][52][53]. Among the other types of locomotion in aerial environments, ballooning has been rarely used [54].

Terrestrial Locomotion
Terrestrial locomotion in animals is mostly achieved by means of legs, thanks to their capability to allow for navigation through rough terrain, obstacles and narrow spaces. Research in walking robots has spanned through the last fifty years [1,55]. The first legged robots appeared in the late 1960s as quadrupeds, whereas biped robots appeared some years later, when, in 1970, Kato started the WABOT project at Waseda University, Tokyo [1,7,56]. They were all characterized by static gaits. Dynamically stable gaits were first approached in the 1980s by Raibert [57], who started with a planar hopping system and elaborated it to a bipedal and quadruped machine. Since the early 1990s, legged robotics moved to the design of more versatile systems able to perform several different operations. This development became possible thanks to the increased available computational power, which allowed for dynamic control of complex systems. Among the successful projects, Honda's "E" and "P" bipedal platform was probably the most famous, and it culminated in the development of the ASIMO robot [58]. Another well-known platform was Boston Dynamics' BigDog, a dynamic quadruped robot that was developed for DARPA in 2006 [59]. BigDog could trot through unstructured terrain and steep slopes and recover from slipping on ice. In the 2010s, Boston Dynamics also released the humanoid Atlas and the smaller quadruped Spot, both capable of performing

Terrestrial Locomotion
Terrestrial locomotion in animals is mostly achieved by means of legs, thanks to their capability to allow for navigation through rough terrain, obstacles and narrow spaces. Research in walking robots has spanned through the last fifty years [1,55]. The first legged robots appeared in the late 1960s as quadrupeds, whereas biped robots appeared some years later, when, in 1970, Kato started the WABOT project at Waseda University, Tokyo [1,7,56]. They were all characterized by static gaits. Dynamically stable gaits were first approached in the 1980s by Raibert [57], who started with a planar hopping system and elaborated it to a bipedal and quadruped machine. Since the early 1990s, legged robotics moved to the design of more versatile systems able to perform several different operations. This development became possible thanks to the increased available computational power, which allowed for dynamic control of complex systems. Among the successful projects, Honda's "E" and "P" bipedal platform was probably the most famous, and it culminated in the development of the ASIMO robot [58]. Another well-known platform was Boston Dynamics' BigDog, a dynamic quadruped robot that was developed for DARPA in 2006 [59]. BigDog could trot through unstructured terrain Robotics 2020, 9, 32 4 of 18 and steep slopes and recover from slipping on ice. In the 2010s, Boston Dynamics also released the humanoid Atlas and the smaller quadruped Spot, both capable of performing complex dynamic tasks (e.g., backflips) [59]. Another humanoid, NAO by Aldebaran Robotics (now SoftBank Robotics), launched in 2008, is now the standard platform for several robotics competitions, such as the RoboCup Standard Platform League [60]. Other example of humanoids are: iCub, for research on cognitive development [61]; WALK-MAN, a rescue robot [62]; Ami, for domotics [63]; REEM-B by PAL-Robotics, for service tasks [64]; LARMbot, for service tasks [65]; ARMAR, for domotics [66,67]. However, some issues still prevent legged machines from being available for large-scale use. The main issue is power source options, since the available ones limit legged robots to very short battery operations, making them suitable only for demonstration. Internal combustion engines are used in some robots, such as BigDog, but they restrict their usage to places where noise and emissions are not a problem. Furthermore, small systems are limited by the performance of the current actuator technology. Biped and quadruped robots are usually extremely expensive, and the task that they can perform can be achieved successfully by simpler and cheaper systems.
Two distinct modes of terrestrial locomotion that use rotation can be found: rolling, when the organism rotates as a whole; and wheels or tracks, where the rotation is limited to an axle or a shaft and to part of the body only. In nature, a few animals are able to roll, while no instances of wheels can be found. On the other hand, wheeled and tracked robot are widespread and represent the most mature mobile robot technology, [1,[68][69][70][71][72][73][74], thanks to their simple control and high performance, despite their limitations with obstacles and rough terrains. Rolling robots have been investigated and many successful prototypes have been proposed, [75][76][77][78][79][80].

Hybrid Locomotion
Hybrid mobile robots are characterized by two or more modes of locomotion. This kind of robots, sometimes called multimodal, can be divided into:

•
Reconfigurable multimodal robots, when they are characterized by two or more locomotion modes performed by the same mechatronic system. The reconfiguration strategy of these class of robots must be carefully studied to reduce "dead" times, when the robot changes from a configuration to another. • Non-reconfigurable multimodal robots, when the different locomotion modes are independent from each other. In this case, it is possible to define hybrid motion strategies when both modes are performed together to optimize navigation and performance.
Moreover, the different modes of locomotion could be part of the same area or could allow the robot to navigate through two or more different environments, as outlined in Figure 1.

Technology Readiness
As mentioned in the introduction, the different modes of locomotion are characterized by a wide range of designs. Some of them have been in use for decades and can be easily controlled even by non-expert users. Others have experienced little success, but their functioning is well understood and documented. In order to estimate the technology level of each mode of locomotion, two different indicators are used in this work. The first one is the Technology Readiness Level (TRL) [107], which is an assessment of the maturity of a particular technology on the following scale: • TRL 1 -basic principles observed • TRL 2 -technology concept formulated • TRL 3 -experimental proof of concept • TRL 4 -technology validated in laboratory • TRL 5 -technology validated in relevant environment • TRL 6 -technology demonstrated in relevant environment • TRL 7 -system prototype demonstration in operational environment • TRL 8 -system complete and qualified • TRL 9 -actual system proven in operational environment.
The second one is the Commercial Readiness Index (CRI) [107], which aims to complement the TRLs by assessing the commercial maturity of technologies across the following levels: An estimation of both TRLs and CRIs of locomotion modes according to the literature presented in this section can be found in Table 1. Only three locomotion modes, namely drones and wheeled/tracked vehicles, are mature enough for commercial applications (CRI 4-6), mainly due to their design being an adaptation of existing vehicles to autonomous or teleoperated navigation. None of them has seen widespread deployment, mainly due to high costs, even if some specific applications are technologically mature (e.g., robotic vacuum cleaners). Several companies are working to develop fully autonomous cars (e.g., Google, Apple), but a combination of high complexity, costs, and legal issue are still relegating them to few prototypes. Most of the bio-inspired designs, on the other hand,

Technology Readiness
As mentioned in the introduction, the different modes of locomotion are characterized by a wide range of designs. Some of them have been in use for decades and can be easily controlled even by non-expert users. Others have experienced little success, but their functioning is well understood and documented. In order to estimate the technology level of each mode of locomotion, two different indicators are used in this work. The first one is the Technology Readiness Level (TRL) [107], which is an assessment of the maturity of a particular technology on the following scale: The second one is the Commercial Readiness Index (CRI) [107], which aims to complement the TRLs by assessing the commercial maturity of technologies across the following levels: Level 5 -Market competition driving widespread deployment • Level 6 -"Bankable" grade asset class.
An estimation of both TRLs and CRIs of locomotion modes according to the literature presented in this section can be found in Table 1. Only three locomotion modes, namely drones and wheeled/tracked vehicles, are mature enough for commercial applications (CRI 4-6), mainly due to their design being an adaptation of existing vehicles to autonomous or teleoperated navigation. None of them has seen widespread deployment, mainly due to high costs, even if some specific applications are technologically mature (e.g., robotic vacuum cleaners). Several companies are working to develop fully autonomous cars (e.g., Google, Apple), but a combination of high complexity, costs, and legal issue are still relegating them to few prototypes. Most of the bio-inspired designs, on the other hand, are characterized by a very low TRL, with widespread research on them but no commercial application and very few functional prototypes. However, the surge of interest in soft and biomimetic robots could result in a spike in their TRL in the next decade.

Terrestrial Hybrid Robots
The wide range of different terrestrial environments, with a huge variety of terrain and conditions, often requires more than a single locomotion mode to achieve a task. Even a simple linear straight motion can be deeply affected by a change of terrain, such as moving from concrete to sand. For these reason, many multimodal land robots were designed. Among them, the most common type is characterized by having both wheel and legs, as per examples in Figures 2 and 3.
Wheels are the most efficient locomotion mode, but they have issues when navigating in rough terrain and they cannot overcome obstacles higher than their radius. Legged locomotion allows robots to move in unstructured terrain, with uneven surfaces, steps and obstacles. Furthermore, wheels and tracks erode the terrain they move in, while walking robots reduce this damage considerably. Both locomotion modes are easy to control when more than three legs and wheels are used, as they ensure static balance. For these reasons, many hybrid robots with legs and wheels have been developed, [108][109][110][111][112][113][114][115][116][117]. Most of them are characterized by independent locomotory systems, such as the examples in Figure 2, with the wheels that can also act as foot for the legged locomotion mode. However, dependent leg-wheel locomotion can be found in robots with wheels that transform into legs to change mode, as shown by the prototypes in Figure 3.
Even though wheeled-legged robots are the majority of terrestrial hybrid robots, some other locomotory mode combinations can be found. MorpHex, in Figure 4a [118], is a reconfigurable hybrid robot that can move either by rolling or as a hexapod. A reconfigurable snake/biped robot is shown in Figure 4b [119], while the robot in [120] combines omni-directional wheels and slithering.
Overall, the main trend in terrestrial hybrid robot can be identified in having a main motion mode with high efficiency and performance (usually rotational) and a secondary one that is used to overcome the obstacles and terrains where the main mode struggles. Thus, the mechanical architecture of most of these robots can be summarized as one of the following two designs: are characterized by a very low TRL, with widespread research on them but no commercial application and very few functional prototypes. However, the surge of interest in soft and biomimetic robots could result in a spike in their TRL in the next decade.

Terrestrial Hybrid Robots
The wide range of different terrestrial environments, with a huge variety of terrain and conditions, often requires more than a single locomotion mode to achieve a task. Even a simple linear straight motion can be deeply affected by a change of terrain, such as moving from concrete to sand. For these reason, many multimodal land robots were designed. Among them, the most common type is characterized by having both wheel and legs, as per examples in Figures 2 and 3.
Wheels are the most efficient locomotion mode, but they have issues when navigating in rough terrain and they cannot overcome obstacles higher than their radius. Legged locomotion allows robots to move in unstructured terrain, with uneven surfaces, steps and obstacles. Furthermore, wheels and tracks erode the terrain they move in, while walking robots reduce this damage considerably. Both locomotion modes are easy to control when more than three legs and wheels are used, as they ensure static balance. For these reasons, many hybrid robots with legs and wheels have been developed, [108][109][110][111][112][113][114][115][116][117]. Most of them are characterized by independent locomotory systems, such as the examples in Figure 2, with the wheels that can also act as foot for the legged locomotion mode. However, dependent leg-wheel locomotion can be found in robots with wheels that transform into legs to change mode, as shown by the prototypes in Figure 3.
Even though wheeled-legged robots are the majority of terrestrial hybrid robots, some other locomotory mode combinations can be found. MorpHex, in Figure 4a [118], is a reconfigurable hybrid robot that can move either by rolling or as a hexapod. A reconfigurable snake/biped robot is shown in Figure 4b [119], while the robot in [120] combines omni-directional wheels and slithering.
Overall, the main trend in terrestrial hybrid robot can be identified in having a main motion mode with high efficiency and performance (usually rotational) and a secondary one that is used to overcome the obstacles and terrains where the main mode struggles. Thus, the mechanical architecture of most of these robots can be summarized as one of the following two designs: • Serial leg structure with a wheel as part of the foot, such as the examples in Figure 2.
• Wheel made of an articulated structure that can reconfigure as a serial leg, as shown in the examples in Figure 3. Terrestrial hybrid robots are usually characterized by static balance and the main research challenge is the motion strategy optimization.         Terrestrial hybrid robots are usually characterized by static balance and the main research challenge is the motion strategy optimization.

Amphibious Hybrid Robots
Amphibious robots can navigate through terrestrial and aquatic environments, and they can be divided into two different subclasses: the first one includes all those robots that use the same locomotion mode to move both through aquatic and terrestrial terrains, such as the examples in Figure 5 [121][122][123][124][125]; the second one includes those robots that use distinct locomotion modes for land and water, such as in Figure 6 [126][127][128]. Many terrestrial hybrid robots were characterized by main and secondary locomotion modes, whereas amphibious robot research aims to have a single locomotion mode (namely legs, wheels or a combination of the two) that can be adapted to underwater motion by including a thrust-generating element. Surprisingly a few amphibious robots evolved from amphibious vehicles such as ATVs and hovercrafts, especially when considering how the most successful and technologically advanced mobile robots with a single locomotion mode were created by embedding autonomous drive into pre-existing vehicle designs.
Overall, the most successful mechanical solution for amphibious hybrid robots is a reconfigurable fin that doubles up as a leg for terrestrial locomotion, such as the one in Figure 5a. Most of these robots employ hexapod walking patterns with static balance given by three legs in contact with the ground at any time, while all six fins are used contemporaneously during aquatic motion. Given the variety of designs, however, it is possible to identify several application-based research challenges, but no single one is shared by all the amphibious robots.

Amphibious Hybrid Robots
Amphibious robots can navigate through terrestrial and aquatic environments, and they can be divided into two different subclasses: the first one includes all those robots that use the same locomotion mode to move both through aquatic and terrestrial terrains, such as the examples in Figure 5 [121][122][123][124][125]; the second one includes those robots that use distinct locomotion modes for land and water, such as in Figure 6 [126][127][128]. Many terrestrial hybrid robots were characterized by main and secondary locomotion modes, whereas amphibious robot research aims to have a single locomotion mode (namely legs, wheels or a combination of the two) that can be adapted to underwater motion by including a thrust-generating element. Surprisingly a few amphibious robots evolved from amphibious vehicles such as ATVs and hovercrafts, especially when considering how the most successful and technologically advanced mobile robots with a single locomotion mode were created by embedding autonomous drive into pre-existing vehicle designs.
Overall, the most successful mechanical solution for amphibious hybrid robots is a reconfigurable fin that doubles up as a leg for terrestrial locomotion, such as the one in Figure 5a. Most of these robots employ hexapod walking patterns with static balance given by three legs in contact with the ground at any time, while all six fins are used contemporaneously during aquatic motion. Given the variety of designs, however, it is possible to identify several application-based research challenges, but no single one is shared by all the amphibious robots.

Aerial Hybrid Robots
Every flying animal-birds, insects, bats-can perform a kind of legged locomotion (hexapod locomotion for insects, biped for birds, quadruped for bats). However, most of their robotic counterparts are designed to aerial motion only, since the requirements for flight are extremely restrictive and adding a terrestrial locomotion system means increasing complexity and weight. Very

Amphibious Hybrid Robots
Amphibious robots can navigate through terrestrial and aquatic environments, and they can be divided into two different subclasses: the first one includes all those robots that use the same locomotion mode to move both through aquatic and terrestrial terrains, such as the examples in Figure 5 [121][122][123][124][125]; the second one includes those robots that use distinct locomotion modes for land and water, such as in Figure 6 [126][127][128]. Many terrestrial hybrid robots were characterized by main and secondary locomotion modes, whereas amphibious robot research aims to have a single locomotion mode (namely legs, wheels or a combination of the two) that can be adapted to underwater motion by including a thrust-generating element. Surprisingly a few amphibious robots evolved from amphibious vehicles such as ATVs and hovercrafts, especially when considering how the most successful and technologically advanced mobile robots with a single locomotion mode were created by embedding autonomous drive into pre-existing vehicle designs.
Overall, the most successful mechanical solution for amphibious hybrid robots is a reconfigurable fin that doubles up as a leg for terrestrial locomotion, such as the one in Figure 5a. Most of these robots employ hexapod walking patterns with static balance given by three legs in contact with the ground at any time, while all six fins are used contemporaneously during aquatic motion. Given the variety of designs, however, it is possible to identify several application-based research challenges, but no single one is shared by all the amphibious robots.

Aerial Hybrid Robots
Every flying animal-birds, insects, bats-can perform a kind of legged locomotion (hexapod locomotion for insects, biped for birds, quadruped for bats). However, most of their robotic counterparts are designed to aerial motion only, since the requirements for flight are extremely restrictive and adding a terrestrial locomotion system means increasing complexity and weight. Very

Aerial Hybrid Robots
Every flying animal-birds, insects, bats-can perform a kind of legged locomotion (hexapod locomotion for insects, biped for birds, quadruped for bats). However, most of their robotic counterparts are designed to aerial motion only, since the requirements for flight are extremely restrictive and adding a terrestrial locomotion system means increasing complexity and weight. Very few research groups have approached this research topic. Among them, Floreano's group developed a bat-inspired robot that is able to fly and crawl thanks to a reconfigurable folding wing design, [129,130]. A few other examples combine drone architectures with legs and wheels, Figure 7, [131][132][133].
Research on hybrid robots that can move through aerial and aquatic environments has been extremely scattered and sparse, with some studies up to the conceptual design stage [134][135][136] but no prototypes yet. The advantages of these robots are outweighed by their rather complex design, having to consider locomotory systems able to generate aerodynamic and fluid-dynamic thrust and lift forces, as well as function and weather extreme environments (underwater currents, wind). Furthermore, their application is limited to few, very specific tasks that can usually be achieved by separate systems. For these reasons, this kind of hybrid robots will difficulty be developed without further advances in pure aerial or aquatic autonomous systems. Overall, this field of research is almost unrepresented, while offering several design advantages when compared to purely flying or land robots, as highlighted by the HeritageBot case study.
The main mechanical challenge in aerial hybrid robots is the weight of the structure. Adding a mechatronic system for land locomotion to a flying robot significantly increases its weight, limiting the available solutions to miniaturized robots (Figure 7) or flying solutions that are oversized when compared to the land locomotion system. Furthermore, a main research gap can be identified in the interaction between the terrestrial and aerial locomotion systems, as further detailed in the HeritageBot case study. few research groups have approached this research topic. Among them, Floreano's group developed a bat-inspired robot that is able to fly and crawl thanks to a reconfigurable folding wing design, [129][130]. A few other examples combine drone architectures with legs and wheels, Figure 7, [131][132][133].
Research on hybrid robots that can move through aerial and aquatic environments has been extremely scattered and sparse, with some studies up to the conceptual design stage [134][135][136] but no prototypes yet. The advantages of these robots are outweighed by their rather complex design, having to consider locomotory systems able to generate aerodynamic and fluid-dynamic thrust and lift forces, as well as function and weather extreme environments (underwater currents, wind). Furthermore, their application is limited to few, very specific tasks that can usually be achieved by separate systems. For these reasons, this kind of hybrid robots will difficulty be developed without further advances in pure aerial or aquatic autonomous systems. Overall, this field of research is almost unrepresented, while offering several design advantages when compared to purely flying or land robots, as highlighted by the HeritageBot case study.
The main mechanical challenge in aerial hybrid robots is the weight of the structure. Adding a mechatronic system for land locomotion to a flying robot significantly increases its weight, limiting the available solutions to miniaturized robots (Figure 7) or flying solutions that are oversized when compared to the land locomotion system. Furthermore, a main research gap can be identified in the interaction between the terrestrial and aerial locomotion systems, as further detailed in the HeritageBot case study.

Case Study: HeritageBot
In this section, a case study of a hybrid walking-flying robot, HeritageBot, is presented. HeritageBot, shown in Figure 8, was developed as part of the HeritageBot project at University of Cassino, with the aim of enhancing the exploration and conservation of Cultural Heritage. The main challenge of this project was to address the survey, inspection, and investigation of archaeological areas that cannot be accessed by human operators, either for safety reasons, inaccessible and/or hazardous environments or to prevent or mitigate the risk of damaging or deteriorating features such as mosaics and floor frescos. The best solution to achieve the task was identified in a flying/walking robot, as documented in [137]. The technical requirements for the platform were defined as the capability of acquire data through tele-operated and autonomous action, aerial and terrestrial mobility, capability of small-scale intervention and collection of samples, and a modular design to embed a variety of sensors according to the kind of necessary intervention.
The advantages of flying and walking robots is analyzed in detail in [138], where the authors describe the development of two complementary robots, a walking platform and a drone. The drone is able to fly over the target area in order to map the environment and localize the main points that need further investigation or intervention. The aerial data is shared with the walking robot, which then walks to the point and performs the tasks, thanks to its better land mobility, battery life, and payload capability. However, while [138] defines a "virtual" framework for the collaboration and usage of flying and walking locomotory systems, two distinct physical systems are employed. In

Case Study: HeritageBot
In this section, a case study of a hybrid walking-flying robot, HeritageBot, is presented. HeritageBot, shown in Figure 8, was developed as part of the HeritageBot project at University of Cassino, with the aim of enhancing the exploration and conservation of Cultural Heritage. The main challenge of this project was to address the survey, inspection, and investigation of archaeological areas that cannot be accessed by human operators, either for safety reasons, inaccessible and/or hazardous environments or to prevent or mitigate the risk of damaging or deteriorating features such as mosaics and floor frescos. The best solution to achieve the task was identified in a flying/walking robot, as documented in [137]. The technical requirements for the platform were defined as the capability of acquire data through tele-operated and autonomous action, aerial and terrestrial mobility, capability of small-scale intervention and collection of samples, and a modular design to embed a variety of sensors according to the kind of necessary intervention. a mosaic). In these cases, both locomotion modes are required for a single system to move onto the target point. Even if tracks are often the preferred locomotion mode for rescue and exploration robots, the risk of damaging precious surfaces requires a legged system. Moreover, legs provide superior stability on rough and uneven terrain, when compared to wheeled or tracked designs. Thus, a legged design was chosen for the walking mode, while a quadcopter architecture allowed HeritageBot to fly. Further details can be found in [139][140][141][142]. Examples of HeritageBot operation are shown in Figures 9  and 10, with the prototype climbing steps and flying, respectively.  Step-climbing operation of the HeritageBot Prototype.
While flying, HeritageBot can easily overcome most of the obstacles and aerial mapping of the target Cultural Heritage area is possible. However, flight is characterized by a short battery life, and to optimize power consumption, walking is performed as the main mode of locomotion. Furthermore, the quadruped stance allows for stable image acquisition. Since the two locomotory systems are independent, hybrid motion strategies can be defined for HeritageBot as follows: • Optimal navigation: instead of planning HeritageBot path from one waypoint to the following one by considering a single locomotion mode, an energy cost function can be defined to take into account the energy consumption of the path by using: flight only; walking only; a combination of the two. Thus, an optimal path can be defined as the one with the minimal energy cost. Even if walking has a lower cost overall, an aerial path can be more convenient (or even the only possible way) in case of large obstacles on the way. • Dynamic balancing: the helices can be used to generate aerodynamic lift to balance the quadruped robot in case of slipping or similar losses of balance (e.g., landslides on the ground where the robot is walking). Moreover, an aerodynamic drag force can be generated to help the robot "walk" on surfaces with high inclination by keeping the feet contacting the ground. • Improved payload: in case of intervention and sample collection, an aerodynamic lift can be used to increase the payload of the system without overloading the linear actuators of the legs.
These three examples show that hybrid robots with independent locomotory systems can be used not only to perform the same tasks of traditional mobile robots, but that the collaboration The advantages of flying and walking robots is analyzed in detail in [138], where the authors describe the development of two complementary robots, a walking platform and a drone. The drone is able to fly over the target area in order to map the environment and localize the main points that need further investigation or intervention. The aerial data is shared with the walking robot, which then walks to the point and performs the tasks, thanks to its better land mobility, battery life, and payload capability. However, while [138] defines a "virtual" framework for the collaboration and usage of flying and walking locomotory systems, two distinct physical systems are employed. In Cultural Heritage, some obstacles could be unsurmountable for walking robots (e.g., a fallen wall, or a mosaic). In these cases, both locomotion modes are required for a single system to move onto the target point. Even if tracks are often the preferred locomotion mode for rescue and exploration robots, the risk of damaging precious surfaces requires a legged system. Moreover, legs provide superior stability on rough and uneven terrain, when compared to wheeled or tracked designs. Thus, a legged design was chosen for the walking mode, while a quadcopter architecture allowed HeritageBot to fly. Further details can be found in [139][140][141][142]. Examples of HeritageBot operation are shown in Figures 9 and 10, with the prototype climbing steps and flying, respectively. Cultural Heritage, some obstacles could be unsurmountable for walking robots (e.g., a fallen wall, or a mosaic). In these cases, both locomotion modes are required for a single system to move onto the target point. Even if tracks are often the preferred locomotion mode for rescue and exploration robots, the risk of damaging precious surfaces requires a legged system. Moreover, legs provide superior stability on rough and uneven terrain, when compared to wheeled or tracked designs. Thus, a legged design was chosen for the walking mode, while a quadcopter architecture allowed HeritageBot to fly. Further details can be found in [139][140][141][142]. Examples of HeritageBot operation are shown in Figures 9  and 10, with the prototype climbing steps and flying, respectively.  Step-climbing operation of the HeritageBot Prototype.
While flying, HeritageBot can easily overcome most of the obstacles and aerial mapping of the target Cultural Heritage area is possible. However, flight is characterized by a short battery life, and to optimize power consumption, walking is performed as the main mode of locomotion. Furthermore, the quadruped stance allows for stable image acquisition. Since the two locomotory systems are independent, hybrid motion strategies can be defined for HeritageBot as follows: • Optimal navigation: instead of planning HeritageBot path from one waypoint to the following one by considering a single locomotion mode, an energy cost function can be defined to take into account the energy consumption of the path by using: flight only; walking only; a combination of the two. Thus, an optimal path can be defined as the one with the minimal energy cost. Even if walking has a lower cost overall, an aerial path can be more convenient (or even the only possible way) in case of large obstacles on the way. • Dynamic balancing: the helices can be used to generate aerodynamic lift to balance the quadruped robot in case of slipping or similar losses of balance (e.g., landslides on the ground where the robot is walking). Moreover, an aerodynamic drag force can be generated to help the robot "walk" on surfaces with high inclination by keeping the feet contacting the ground. • Improved payload: in case of intervention and sample collection, an aerodynamic lift can be used to increase the payload of the system without overloading the linear actuators of the legs.
These three examples show that hybrid robots with independent locomotory systems can be Step-climbing operation of the HeritageBot Prototype.
While flying, HeritageBot can easily overcome most of the obstacles and aerial mapping of the target Cultural Heritage area is possible. However, flight is characterized by a short battery life, and to optimize power consumption, walking is performed as the main mode of locomotion. Furthermore, the quadruped stance allows for stable image acquisition. Since the two locomotory systems are independent, hybrid motion strategies can be defined for HeritageBot as follows: • Optimal navigation: instead of planning HeritageBot path from one waypoint to the following one by considering a single locomotion mode, an energy cost function can be defined to take into account the energy consumption of the path by using: flight only; walking only; a combination of the two. Thus, an optimal path can be defined as the one with the minimal energy cost. Even if walking has a lower cost overall, an aerial path can be more convenient (or even the only possible way) in case of large obstacles on the way.

•
Dynamic balancing: the helices can be used to generate aerodynamic lift to balance the quadruped robot in case of slipping or similar losses of balance (e.g., landslides on the ground where the robot is walking). Moreover, an aerodynamic drag force can be generated to help the robot "walk" on surfaces with high inclination by keeping the feet contacting the ground. • Improved payload: in case of intervention and sample collection, an aerodynamic lift can be used to increase the payload of the system without overloading the linear actuators of the legs.
These three examples show that hybrid robots with independent locomotory systems can be used not only to perform the same tasks of traditional mobile robots, but that the collaboration between multiple locomotion modes could push the capabilities of the robot over some of the traditional limits.

Discussion
Research on mobile robots spanned over 60 years, whereas hybrid designs with two locomotion modes stretch over the last two decades only, or the last one without considering wheeled/legged robots. A single example [136] that can navigate through more than two environments has been identified. Two distinct design trends emerge from the literature on hybrid robots: • Main and secondary modes of locomotion: one of the main design trends behind hybrid mobile robot is to have a primary mode of locomotion that is usually more efficient or performing, and a secondary mode that is used only to overcome obstacles. An example of this design is given by wheeled/legged mobile robots. • Reconfigurable mode of locomotion: many hybrid robots use a single mode of locomotion (e.g., legs, or helices) that works in more than one environment with minimal change. For examples, helices can be used to generate thrust in water, and lift for flying, or legs can swim and walk.
The main research gaps and issues in hybrid mobile robots can be identified as: • Independent locomotion modes: robots with two or more independent locomotory systems are usually either designed to use them one at a time. Little research has been done on how multiple modes can interact and collaborate. • Applications: while the most common designs are usually task-oriented, many bio-inspired robots are developed without a target application, but only to prove the feasibility of new designs (e.g., most of the flapping wings robot). Even if those robots are interesting, the lack of focus limits them to low TRLs.

Discussion
Research on mobile robots spanned over 60 years, whereas hybrid designs with two locomotion modes stretch over the last two decades only, or the last one without considering wheeled/legged robots. A single example [136] that can navigate through more than two environments has been identified. Two distinct design trends emerge from the literature on hybrid robots:

•
Main and secondary modes of locomotion: one of the main design trends behind hybrid mobile robot is to have a primary mode of locomotion that is usually more efficient or performing, and a secondary mode that is used only to overcome obstacles. An example of this design is given by wheeled/legged mobile robots.

•
Reconfigurable mode of locomotion: many hybrid robots use a single mode of locomotion (e.g., legs, or helices) that works in more than one environment with minimal change. For examples, helices can be used to generate thrust in water, and lift for flying, or legs can swim and walk.
The main research gaps and issues in hybrid mobile robots can be identified as: • Independent locomotion modes: robots with two or more independent locomotory systems are usually either designed to use them one at a time. Little research has been done on how multiple modes can interact and collaborate.

•
Applications: while the most common designs are usually task-oriented, many bio-inspired robots are developed without a target application, but only to prove the feasibility of new designs (e.g., most of the flapping wings robot). Even if those robots are interesting, the lack of focus limits them to low TRLs. • Performance index: a general index to represent the performance of the robot, as for example energy consumption and efficiency during different motion modes, is not available. Thus, it is difficult to compare robots with different modes of motion even when they move in similar environments.
Overall, the mechanical design of traditional mobile robots is secondary to the development of their control and navigation system, since the technology in mechatronics is more mature than the current AI and navigation systems. However, the fast development and widespread availability of new control and navigation technologies in the 2000s and 2010s is leading to a point where the mechanical design of hybrid robots becomes the limiting factor to commercial success or widespread availability. Even nowadays, research on multimodal robots is focused on the mechanical system itself and on low-level control rather than high-level control and navigation, especially in the case of reconfigurable robots.

Conclusions
This paper summarized the current design solutions for mobile robots in any environment, with an emphasis on hybrid mobility combining capabilities for terrestrial, aerial, and aquatic locomotion. A survey was presented to discuss solutions and characteristics of mobile robots, in order to introduce hybrid designs that integrate different technologies for different environments. The HeritageBot Platform was presented as a case study from the direct experiences of the authors to show the successful integration of mobile modules into a robot with hybrid synergic capabilities.