Unverified50% confidenceFactExact time
测量协方差矩阵 R 越大滤波器越不信任该传感器读数，越小则越快响应观测变化
1
Sources
50%
Confidence
Long-term
Relevance
7/12/2026
First Seen
Sources
Multi-Sensor Kalman Fusion in Practice: A Complete Breakdown of a Drone Tracking System
redditr/learnmachinelearning7/11/2026
Related Claims
Unverified为每个传感器独立建模测量协方差矩阵R是多传感器融合区别于单传感器扩展的本质，R设置过小会导致过度信任该传感器测量值78% similarUnverified当标注不一致存在时，增加更多数据只会成倍放大混乱，向模型提供自相矛盾的监督信号67% similarUnverifiedQuantization error has a non-uniform distribution characteristic; the Q/K/V matrices of the attention mechanism and the last few layers are more sensitive to quantization.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/490365API
curl https://kongchang.com/api/v1/knowledge/claims/490365MCP
get_claim(id=490365)