A Multi-Sensor-Based Terrain Perception Model for Locomotion Selection of Hybrid Mobile Robots

Autonomous mobile robots can travel and perform activities without human involvement. These robots are outfitted with various sensors, actuators, and onboard computing capabilities to detect their environment, make decisions, and execute tasks independently. The performance of such robots depends on how well they perceive their environments. Most sensors are employed towards the surroundings for decision-making and path planning. This paper tries a different approach that complements environment perception to increase the robot’s perception ability. The suggested approach uses standard sensors like cameras, light sensors, distance sensors, accelerometers, and gyroscope sensors in a compact sensor box meant to face the ground to provide terrain perception ability to the robots. The sensor box uses a supervised learning-based concept to categorize different types of terrain and estimate how practicable they are in terms of percentage, i.e., the terrain perception percentage. The experimental outcomes depict that the terrain perception model gives a reliable indication of the practicability of terrains, thereby increasing the perception abilities of robots.