An In-Memory Physics Environment as a World Model for Robot Motion Planning

Abstract

This paper investigates the possibility of utilizing a physics simulation environment as the imagination of a robot, where it creates a replica of the detected terrain in a physics simulation environment in its memory, and “imagines” a simulated version of itself in that memory, performing actions and navigation on the terrain. The physics of the environment simulates the movement of robot parts and its interaction with the objects in the environment and the terrain, thus avoiding the need for explicitly programming many calculations. The robot chooses the best possible action from multiple simulations of movement, and executes it in the real world. Moreover, as the complexity of motion increases with each degree of freedom of the robot’s joints, this paper also explores the utility of uniform pseudo-randomness to explore the fitness landscape of robot motility, and compares it with Computational Intelligence algorithms. Such techniques could potentially simplify the algorithmic complexity of programming multi-jointed robots, and also be capable of dynamically adjusting the “mental” simulation of the robot when it encounters environments with different gravity, viscosity or traction, merely by adjusting parameters of the simulated environment.