EV-Planner: Energy-Efficient Robot Navigation via Event-Based Physics-Guided Neuromorphic Planner

TL;DR

EV-Planner addresses energy-efficient autonomous drone navigation through moving gates by fusing neuromorphic event-based perception with physics-guided planning. It employs a shallow SNN for fast, unsupervised object detection from event streams and a PgNN trained on motor dynamics to predict near-minimum-energy flight times, all coordinated via a ROS-based symbolic planner. The approach demonstrates substantial actuation-energy savings and robust collision-free trajectories in Gazebo/RotorS simulations, supported by ablation studies and comparisons to depth-based baselines and SOTA event-perception methods. This sensor-fused, neuro-symbolic framework advances energy-efficient planning for small drones and offers a practical path toward real-world deployment with interpretable physics priors.

Abstract

Vision-based object tracking is an essential precursor to performing autonomous aerial navigation in order to avoid obstacles. Biologically inspired neuromorphic event cameras are emerging as a powerful alternative to frame-based cameras, due to their ability to asynchronously detect varying intensities (even in poor lighting conditions), high dynamic range, and robustness to motion blur. Spiking neural networks (SNNs) have gained traction for processing events asynchronously in an energy-efficient manner. On the other hand, physics-based artificial intelligence (AI) has gained prominence recently, as they enable embedding system knowledge via physical modeling inside traditional analog neural networks (ANNs). In this letter, we present an event-based physics-guided neuromorphic planner (EV-Planner) to perform obstacle avoidance using neuromorphic event cameras and physics-based AI. We consider the task of autonomous drone navigation where the mission is to detect moving gates and fly through them while avoiding a collision. We use event cameras to perform object detection using a shallow spiking neural network in an unsupervised fashion. Utilizing the physical equations of the brushless DC motors present in the drone rotors, we train a lightweight energy-aware physics-guided neural network (PgNN) with depth inputs. This predicts the optimal flight time responsible for generating near-minimum energy paths. We spawn the drone in the Gazebo simulator and implement a sensor-fused vision-to-planning neuro-symbolic framework using Robot Operating System (ROS). Simulation results for safe collision-free flight trajectories are presented with performance analysis, ablation study and potential future research directions

Figures (12)

• Figure 1: High-level overview of EV-Planner framework
• Figure 2: On the left, the spikes generated do not exceed the membrane potential threshold due to the absence of enough events. On the right, due to object motion, densely generated events produce more input spikes which exceed the membrane potential threshold to generate output spikes.
• Figure 3: Spike-based neural architecture responsible for filtering events from objects based on speed. This SNN is used for subsequent object tracking.
• Figure 4: Isolating the moving object and finding the center of the bounding box in the pixel coordinate system at different depths.
• Figure 5: Energy model of brushless DC motor of quadrotor propellers