Legged robots are mainly designed to traverse unstructured environments where wheeled robots have limited mobility. Their applications range from nuclear decommissioning to mining, search and rescue, inspection and surveillance. In addition, they can be applied to flank human workers in order to reduce labor accidents, as well as in elderly care.
The next generation of legged robots are envisioned to operate either autonomously or semi-autonomously (through tele-operation) over uneven terrains. This requires the rejection and compensation of disturbances, the exploitation of visual feedback, and the ability to manipulate both fragile and heavy objects. The main ingredients for legged locomotion are planning, control, perception and state estimation. Thanks to recent advances, robots can acquire online the 3D map of the terrain (exteroceptive sensing), which can be useful both for inspection and to enhance locomotion (e.g. for foothold selection). At a higher level, navigation algorithms can be developed to avoid (fixed or moving) obstacles, control the locomotion speed and attain a specific position (goal).
The substantial gap between simulation and reality is due to a number of different factors that can make the actual robot state diverge from the original plan. The source of errors can be due to: tracking delays in the controller, sensor calibration errors, filtering delays, inaccuracy in the 3D map, unforeseen events (external pushes, slippages), or dynamically changing and deformable terrains (e.g. rolling stones, mud). Indeed, the stones the robot is stepping onto can collapse, move, roll away, resulting in a (possibly catastrophic) loss of balance. Other sources of error include: structural compliance and modeling errors in general. Re-planning is a crucial feature to intrinsically cope with the problem of error accumulation in real scenarios. It is a mechanism to adapt to the terrain, reject disturbances,
while, at the same time, promptly follow desired user commands.
Heuristic strategies have been proven to be useful for decades in locomotion. However, when the complexity of the terrain to be negotiated increases, or when the execution of the requested task involves highly dynamic motions, optimization-based planning strategies are preferable. Indeed, they enable robots to reach performance limits and achieve agile maneuvers. Being able to merge re-planning and optimization during locomotion (i.e. online) is still an open research problem because of the high computational complexity. Many studies focus on developing simplified models to reduce the complexity of the optimization, trading off accuracy with computational efficiency. Moreover, an effective locomotion framework encompasses different levels of autonomy. At the low level, we have locomotion strategies (e.g. different gaits) that aim to balance the robot while dealing with the terrain, but require high-level commands (e.g. desired velocities). At the high level, the focus is more on improving robot autonomy by orchestrating the locomotion strategies in order to fulfill some user requirements (e.g., reaching a desired location, or opening a door). These algorithms choose the most suitable locomotion strategy and adapt the gait parameters (e.g., according to the terrain difficulty) to accomplish the given task (e.g., reaching a goal, picking up an object). This level takes also care of ensuring smooth transitions between different tasks.
Further scenarios, include robotic systems in cluttered environments. This can require specific strategies that involve whole-body loco-manipulation while exploiting multiple contacts and generate collision-free trajectories. Against this background, this Research Topic solicits research articles on recent results in
legged robotics, including, but not limited to:
• Navigation and obstacle avoidance.
• Online optimization-based re-planning (e.g., to compensate for tracking, modeling, estimation errors).
• Planning of agile non periodic maneuvers (e.g., jumps on high obstacles).
• Reactive strategies (e.g., reflexes).
• Robust planning considering model uncertainties.
• Model-based whole-body control with task priorities.
• Detection, estimation, compensation of external disturbances (e.g., opening a self-closing door with a spring-hinge).
• Loco-manipulation: having a mobile platform (e.g. tracked, legged or wheeled) and a manipulator on top of it (centaur) opens the field to study whole-body reachability exploiting the motion of the mobile platform.
• Locomotion on deformable or dynamically changing terrains. Locomotion strategies that can deal with terrain that can change: moderately (e.g., grass, sand, mud, gravel) or abruptly (e.g., rolling stones).
• Patrolling, surveillance, inspection (e.g., acoustics, gas detection, radiation sensors).
• Walking inside cluttered environments and narrow passages. This involve studies on whole body controller and planners that exploit multiple contacts (e.g. in different points of the body than just hand/feet) and avoid collisions with the environment.
• Datasets and template terrain for benchmarking locomotion algorithms.
Legged robots are mainly designed to traverse unstructured environments where wheeled robots have limited mobility. Their applications range from nuclear decommissioning to mining, search and rescue, inspection and surveillance. In addition, they can be applied to flank human workers in order to reduce labor accidents, as well as in elderly care.
The next generation of legged robots are envisioned to operate either autonomously or semi-autonomously (through tele-operation) over uneven terrains. This requires the rejection and compensation of disturbances, the exploitation of visual feedback, and the ability to manipulate both fragile and heavy objects. The main ingredients for legged locomotion are planning, control, perception and state estimation. Thanks to recent advances, robots can acquire online the 3D map of the terrain (exteroceptive sensing), which can be useful both for inspection and to enhance locomotion (e.g. for foothold selection). At a higher level, navigation algorithms can be developed to avoid (fixed or moving) obstacles, control the locomotion speed and attain a specific position (goal).
The substantial gap between simulation and reality is due to a number of different factors that can make the actual robot state diverge from the original plan. The source of errors can be due to: tracking delays in the controller, sensor calibration errors, filtering delays, inaccuracy in the 3D map, unforeseen events (external pushes, slippages), or dynamically changing and deformable terrains (e.g. rolling stones, mud). Indeed, the stones the robot is stepping onto can collapse, move, roll away, resulting in a (possibly catastrophic) loss of balance. Other sources of error include: structural compliance and modeling errors in general. Re-planning is a crucial feature to intrinsically cope with the problem of error accumulation in real scenarios. It is a mechanism to adapt to the terrain, reject disturbances,
while, at the same time, promptly follow desired user commands.
Heuristic strategies have been proven to be useful for decades in locomotion. However, when the complexity of the terrain to be negotiated increases, or when the execution of the requested task involves highly dynamic motions, optimization-based planning strategies are preferable. Indeed, they enable robots to reach performance limits and achieve agile maneuvers. Being able to merge re-planning and optimization during locomotion (i.e. online) is still an open research problem because of the high computational complexity. Many studies focus on developing simplified models to reduce the complexity of the optimization, trading off accuracy with computational efficiency. Moreover, an effective locomotion framework encompasses different levels of autonomy. At the low level, we have locomotion strategies (e.g. different gaits) that aim to balance the robot while dealing with the terrain, but require high-level commands (e.g. desired velocities). At the high level, the focus is more on improving robot autonomy by orchestrating the locomotion strategies in order to fulfill some user requirements (e.g., reaching a desired location, or opening a door). These algorithms choose the most suitable locomotion strategy and adapt the gait parameters (e.g., according to the terrain difficulty) to accomplish the given task (e.g., reaching a goal, picking up an object). This level takes also care of ensuring smooth transitions between different tasks.
Further scenarios, include robotic systems in cluttered environments. This can require specific strategies that involve whole-body loco-manipulation while exploiting multiple contacts and generate collision-free trajectories. Against this background, this Research Topic solicits research articles on recent results in
legged robotics, including, but not limited to:
• Navigation and obstacle avoidance.
• Online optimization-based re-planning (e.g., to compensate for tracking, modeling, estimation errors).
• Planning of agile non periodic maneuvers (e.g., jumps on high obstacles).
• Reactive strategies (e.g., reflexes).
• Robust planning considering model uncertainties.
• Model-based whole-body control with task priorities.
• Detection, estimation, compensation of external disturbances (e.g., opening a self-closing door with a spring-hinge).
• Loco-manipulation: having a mobile platform (e.g. tracked, legged or wheeled) and a manipulator on top of it (centaur) opens the field to study whole-body reachability exploiting the motion of the mobile platform.
• Locomotion on deformable or dynamically changing terrains. Locomotion strategies that can deal with terrain that can change: moderately (e.g., grass, sand, mud, gravel) or abruptly (e.g., rolling stones).
• Patrolling, surveillance, inspection (e.g., acoustics, gas detection, radiation sensors).
• Walking inside cluttered environments and narrow passages. This involve studies on whole body controller and planners that exploit multiple contacts (e.g. in different points of the body than just hand/feet) and avoid collisions with the environment.
• Datasets and template terrain for benchmarking locomotion algorithms.