Deep Photonic Reservoir Computer Meets UAV Control: An Ultra-Fast Learning-Based Compensator for Agile Flight in Confined Space
Qinxiao Ma, Ruiqian Li, Cheng Wang, Yang Wang
AI summary
Problem
UAVs operating in confined spaces suffer from performance degradation due to nonlinear, time-varying unmodeled aerodynamic dynamics that traditional controllers cannot capture. Existing learning-based compensators face prohibitive computational costs, vanishing gradients, and heavy reliance on explicit historical data.
Approach
The authors integrate a hardware-based deep photonic reservoir computer with a feedforward PID controller to predict residual aerodynamic forces. By leveraging semiconductor laser dynamics for intrinsic temporal memory, the system only requires training a linear readout layer via ridge regression.
Key results
- Residual-force prediction accuracy matches or exceeds TCN and MLP baselines in CFD simulations
- Training time reduced from tens of minutes to milliseconds via linear ridge regression
- Inference latency reduced to microseconds, enabling real-time control updates
- Framework supports real-time readout reconfiguration for adaptation to unseen flight scenarios
Why it matters
Delivers a lightweight, ultra-fast, and energy-efficient compensation framework that enables stable, agile UAV flight in complex confined environments, demonstrating the practical viability of photonic computing for robotics.
Abstract
Unmanned aerial vehicles (UAVs) operating in confined, cluttered environments face significant performance degradation due to nonlinear, time-varying unmodeled dynam- ics—such as ground/ceiling effects and wake recirculation—that are unaccounted for in traditional controllers. While learning- based compensators (e.g., MLPs, TCNs, LSTMs) struggle with historical data dependency, vanishing gradients, and prohibitive computational costs, this work pioneers the integration of a deep photonic reservoir computer (PRC) with feedforward control to overcome these limitations. Harnessing semicon- ductor laser dynamics and optical feedback, our hardware- implemented deep PRC architecture achieves intrinsic tempo- ral memory without explicit historical inputs, while reducing training time from hours to milliseconds and slashing inference latency to nanoseconds. Reliable high-performance CFD sim- ulations capturing proximity-induced flows demonstrate that deep PRC delivers residual-force prediction accuracy com- parable to or exceeding TCN/MLP baselines, while training only a linear readout layer via ridge regression. By injecting these predictions into a nonlinear feedback PID controller via a feedforward channel, the framework significantly enhances closed-loop tracking stability in confined spaces. Essentially, this work establishes the first deep PRC-based lightweight, ultra- fast solution for real-time UAV dynamic compensation, with promising extensibility to unseen scenarios with more complex fluid environments.