Issue 41 Unmanned Systems Technology December/January 2022 PteroDynamics X-P4 l Sense & avoid l 4Front Robotics Cricket l Autonomous transport l NWFC-1500 fuel cell l DroneX report l OceanScout I Composites I DSEI 2021 report

48 Digest | 4Front Robotics Cricket planning software to solve. The second was the use of a novel search algorithm that combines aspects of systematic sampling, random sampling and gradient ascent search techniques. A novel mapping algorithm combining spatial occupancy mapping and neural networking was the third. The use of tracks on the end of articulated legs is the Cricket’s most unusual characteristic, and one arrived at by considering the pros and cons of the combination of wheels and legs. Dr Ramirez-Serrano emphasises that wheeled-legged robots are agile, versatile and enable rapid, long- distance mobility over challenging horizontal terrains. He adds though that they do not provide the agility and mobility needed to negotiate almost vertical surfaces and other surfaces on which the wheels will require robots to apply “a tremendous number of forces” to maintain stability. As well as enabling robots to move through complex, slippery, rocky, rough or loose sandy terrain that is challenging for wheels, tracks also allow for trajectory optimisation for tracked and/or wheeled quadrupedal robots capable of hybrid walking-driving-climbing locomotion, he says, as well as providing grasping abilities for climbing ladders and navigating inside industrial facilities. “The Cricket specifically addressed the design and operation of a type of hybrid mechanism with four legs and four tracks or wheels that gives the robot the ability to reconfigure itself into different architectures,” he says, “That allows it to traverse and operate inside chaotic, hazardous confined spaces including vertical pipes, manoeuvre through underground pipe and cable galleries, climb unstructured surfaces, and overcome valves, scaffolding structures and so on.” Dr Ramirez-Serrano adds that significant challenges remain in areas including autonomy, locomotion and adaptability before robots can be successfully deployed in such environments. These include whole-body motion planning, sensing and motion estimation. “The challenge in planning for multi- legged robots comes from the problem’s complexity, and often entails a trade-off between planning time, detail of the robot model and the planning horizon.” Dealing with unpredictability He also says the large uncertainties encountered inside collapsed structures and other disaster environments bring control and mobility challenges to robotics, explaining that there are no control recipes that can be applied to these robots in the way that there are for industrial robots. Therefore, the task facing engineers, he says, is to “develop solutions that combine the theory of exact and approximate Markov decision processes with iterative algorithms that are guaranteed to learn an action- value function, enabling robots to better understand their physical space and improve their behaviours within it.” He adds that machine learning techniques need to be improved to guarantee smooth, whole-body motion planning for a collision-free trajectory despite uncertainties in the terrain, the task and the robot’s capabilities. A further challenge is the development of enhanced locomotion algorithms for multi-legged robots tasked with moving over complex granular materials and surfaces such as sand, rocks and mud, Dr Ramirez-Serrano notes. “Despite decades of terramechanics studies of wheeled and tracked vehicles, little is known about how to move legged robots effectively across granular media.” He points out that although there are many technologies that robots can use to sense and monitor their surroundings and their movements, sensors do not provide a full picture, are subject to various errors and cannot predict how the environment will react when interacting with the robot. Dr Ramirez-Serrano first formulated the concept that became the Cricket in 2010. It included the first mechanical design and subsequent mathematical models, control mechanisms and 3D reconfigurable navigation architectures. The robot still remains true to the original concept but has adopted improved compliant actuators, electronics and sensing systems. The Cricket has December/January 2022 | Unmanned Systems Technology The ability to adopt a flat pose gives the Cricket the option of a very low profile to traverse spaces with severe height restrictions