Fuzzy Encoding-Decoding to Improve Spiking Q-Learning Performance in Autonomous Driving
For autonomous driving researchers using spiking neural networks, this work provides an incremental improvement to spiking reinforcement learning by enhancing representation and value estimation.
The paper proposes a fuzzy encoder-decoder architecture to improve spiking Q-learning for autonomous driving, addressing information loss and weak Q-value discrimination. Experiments on HighwayEnv show the method closes the performance gap between spiking and non-spiking multi-modal Q-networks.
This paper develops an end-to-end fuzzy encoder-decoder architecture for enhancing vision-based multi-modal deep spiking Q-networks in autonomous driving. The method addresses two core limitations of spiking reinforcement learning: information loss stemming from the conversion of dense visual inputs into sparse spike trains, and the limited representational capacity of spike-based value functions, which often yields weakly discriminative Q-value estimates. The encoder introduces trainable fuzzy membership functions to generate expressive, population-based spike representations, and the decoder uses a lightweight neural decoder to reconstruct continuous Q-values from spiking outputs. Experiments on the HighwayEnv benchmark show that the proposed architecture substantially improves decision-making accuracy and closes the performance gap between spiking and non-spiking multi-modal Q-networks. The results highlight the potential of this framework for efficient and real-time autonomous driving with spiking neural networks.