summary

• UberによるPrediction framework REINFORCE の提案
  • Mapや交通ルールから好ましい予測のoutputを出したいが、基本的に微分不可能
  • REINFORCE gradient estimatorを使う
• minADEだけ使っても意味がないよ

Background

• 割と重要なことを言ってる

• この論文はMixture of Gaussian model

Method

• Prior
  • Reachable Lanes
  • SDV Route
• Prior-informed loss function

Metrics

• 1. Motion Forecasting
  • (i) the final lane error (trajectory waypoint inside vs. outside the reachable lanes at 5 seconds into the future) to measure map understanding
  • (ii) minimum average displacement error (minADE) to show the recall of our motion forecasts at different time horizons
  • (iii) mean average displacement error (meanADE), idea of the precision of our predictions since unrealistic samples severely harm this metric
• これだけでは足りない
  • 1. recall is easily achievable at the expense of precision by simply predicting very fan-out(四方八方 に散る) distributions.
  • 2. precision in the motion forecasts is critical for safe motion planning
• 2. Ego-Motion Planning
  • We feed the planner with 50 trajectory samples for each vehicle, as a Monte Carlo approximation of the marginal distribution
  • 乗り心地も評価
    • We focus on the safety-related metric of collision rate (% of time the SDV plan collides with any other traffic participant in the ground-truth, for a future horizon of 5 seconds).
    • lateral acceleration and jerk

Experiment

Discussion

• 非常に社会実装のことを考えたPrediction評価