Adaptive Robot Path Planning Using a Spiking Neuron Algorithm With Axonal Delays

Abstract

A path planning algorithm for outdoor robots, which is based on neuronal spike timing, is introduced. The algorithm is inspired by recent experimental evidence for experience-dependent plasticity of axonal conductance. Based on this evidence, we developed a novel learning rule that altered axonal delays corresponding to cost traversals and demonstrated its effectiveness on real-world environmental maps. We implemented the spiking neuron path planning algorithm on an autonomous robot that can adjust its routes depending on the context of the environment. The robot demonstrates the ability to plan different trajectories that exploit smooth roads when energy conservation is advantageous, or plan the shortest path across a grass field when reducing distance traveled is beneficial. Because the algorithm is suitable for spike-based neuromorphic hardware, it has the potential of realizing orders of magnitude gains in power efficiency and computational gains through parallelization.