Robot Obstacle Avoidance Using Dynamic SNN Thresholds
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing spiking neural networks (SNNs) for obstacle avoidance in robots focus solely on synaptic plasticity, ignoring intrinsic plasticity of neurons, which limits their ability to maintain homeostasis and adapt to degraded environments, leading to inefficiencies in energy consumption and navigation.
Innovation Solution
A biologically reasonable dynamic energy-time threshold method is introduced, comprising a dynamic energy threshold and a dynamic time threshold, integrated to maintain homeostasis and enhance the expressive capacity of SNNs, replacing static thresholds with formulas that correlate membrane potential and depolarization rates to adapt to changing environments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If static thresholds are used in SNN, then the model structure is simple, but the ability to maintain homeostasis and adapt to degraded environments is limited
Solution Approach 1:
The patent applies the dynamics principle by replacing static thresholds with dynamic thresholds that adapt over time. Specifically, it introduces a dynamic energy threshold that evolves based on network activity statistics and a dynamic time threshold that adjusts based on spike timing patterns. This allows the SNN to maintain homeostasis and adapt to degraded environments while preserving biological plausibility.
Solution Approach 2:
The patent implements parameter changes by making the threshold parameters dynamic rather than fixed. The energy threshold is updated based on running statistics of membrane potentials and spike rates, while the time threshold adapts based on inter-spike intervals. These parameter changes enable the network to maintain intrinsic plasticity homeostasis without requiring complex architectural modifications.
2Adaptability or versatility
If intrinsic plasticity is ignored in SNN, then the model is simpler to implement, but the expressive capacity and homeostasis maintenance are limited
Solution Approach 1:
The patent applies segmentation by separating intrinsic plasticity into two distinct threshold mechanisms: the dynamic energy threshold that operates on a slower timescale to maintain global homeostasis, and the dynamic time threshold that operates on a faster timescale to capture local temporal patterns. This segmentation allows each mechanism to be implemented independently while working together to enhance expressive capacity.
Solution Approach 2:
The patent implements feedback mechanisms where the dynamic energy threshold is updated based on feedback from network-wide activity statistics (mean membrane potential and coefficient of variation), and the dynamic time threshold is updated based on feedback from local spike timing patterns. This feedback-driven adaptation enables intrinsic plasticity without requiring complex supervisory signals.
3Reliability
If dynamic thresholds with biological background are applied, then homeostasis is maintained better, but the computational overhead increases
Solution Approach 1:
The patent applies self-service by designing threshold update mechanisms that use only information already available in the network during normal operation. The dynamic energy threshold is computed from existing membrane potential and spike rate statistics, while the dynamic time threshold is computed from existing spike timing data. This self-service approach maintains homeostasis without requiring additional computational resources or external control signals.
Data Source
AI summary
A method for obstacle avoidance in degraded environments of robots based on intrinsic plasticity of an SNN is disclosed. A decision network in a synaptic autonomous learning module takes lidar data, distance from a target point and velocity at a previous moment as state input, and outputs the velocity of left and right wheels of the robot through the autonomous adjustment of the dynamic energy-time threshold, so as to carry out autonomous perception and decision making. The method solves the difficulty of the lack of intrinsic plasticity in the SNN, which leads to the difficulty of adapting to degraded environments due to the homeostasis imbalance of the model, is successfully deployed in mobile robots to maintain a stable trigger rate for autonomous navigation and obstacle avoidance in degraded, disturbed and noisy environments, and has validity and applicability on different degraded scenes.