Directly-trained Spiking Neural Networks for Deep Reinforcement Learning: Energy efficient implementation of event-based obstacle avoidance on a neuromorphic accelerator.
Luca ZanattaAlfio Di MauroFrancesco BarchiAndrea BartoliniLuca BeniniAndrea AcquavivaPublished in: Neurocomputing (2023)
Keyphrases
- efficient implementation
- obstacle avoidance
- reinforcement learning
- spiking neural networks
- mobile robot
- path planning
- visually guided
- biologically inspired
- spiking neurons
- learning rules
- motion planning
- state space
- neural network
- hardware implementation
- artificial neural networks
- robot control
- autonomous robots
- real time
- learning algorithm
- feed forward
- expert systems
- dynamic environments
- machine learning
- biologically plausible
- motor control
- multi modal