I Get Around

When it comes to negotiating unstructured environments with highly variable terrains, robots face significant challenges that have yet to be solved. The ability to traverse such terrains is crucial for applications such as search and rescue operations, exploration of unknown areas, and agricultural tasks. Choosing an appropriate robot morphology plays a vital role in determining their effectiveness in navigating these diverse landscapes.

Wheeled robots are a popular choice for relatively flat and smooth terrains. They offer excellent stability and maneuverability, making them suitable for indoor environments, well-paved surfaces, or even outdoor areas with minimal obstacles. However, their effectiveness decreases when faced with uneven or rocky terrains where their wheels may struggle to maintain traction or encounter difficulties in overcoming obstacles.

Drones offer a unique advantage by bypassing many of the challenges associated with ground-based robots. Drones excel in aerial reconnaissance, surveillance, and rapid mapping of expansive or inaccessible areas. They can quickly fly over highly variable terrains, including mountainous regions, dense forests, and disaster-stricken zones, providing valuable real-time data and imagery. However, drones have limited payload capacity, shorter flight durations, and are susceptible to weather conditions such as strong winds, which can hinder their performance in certain scenarios.

Other modes of locomotion have similar trade-offs, which means that no particular style of robot is ideal where the terrain is variable or unknown. Engineers at Caltech's Center for Autonomous Systems and Technologies realized that solutions to this problem already exist in nature. Inspired by animals such as the Chukar and Hoatzin birds that, at times, repurpose their wings as another pair of legs, or use their wings to assist them in walking up steep inclines, they have developed a multi-modal robot that can adapt its means of locomotion to suit the environment it finds itself in.

Called the M4 (Multi-Modal Mobility Morphobot), this robot has four wheels that allow it to roll along like a traditional wheeled robot. But in the center of each wheel is a rotor, and M4 can reposition the wheels so that they are positioned horizontal to the ground. By spinning up those rotors, M4 can then take to the sky in the form of a quadcopter drone. While on the ground, joints on the wheel assemblies enable the robot to move using a walking motion, something like a quadrupedal robot. This wide range of capabilities allows M4 to choose the best, lowest-energy, mode of locomotion for whatever situation it finds itself in. Some other interesting abilities, like crouching or standing on two legs to look over a tall object, are also made possible by this unique design.

Control over the low-level locomotion functions are handled by a pair of microcontrollers — an Arm Cortex-M3-based processor is utilized to control wheeled motion, while a separate Arm Cortex-M7 processor serves as the flight controller. At this time, the path planning algorithms run on a ground station computer. However, the team has been experimenting with the use of an onboard NVIDIA Jetson Nano computer to handle path planning. In conjunction with an Intel RealSense depth camera, this allowed the M4 to operate fully autonomously.

Moving forward, the researchers intend to improve the walking capabilities of the robot, which are presently in the proof of concept stage. They are also investigating options to give M4 advanced capabilities for manipulating external objects. Improvements such as these, along with enhancements to the path planning algorithms, might make M4 a viable option where no robot is presently practical. One day, M4 robots might even explore other planets.