Identification of the Longitudinal Action Required when Approaching Traffic Lights

Benchmark vs. Logistic Regression, Random Forest, Boosted Tree & Neural Network

Design an AI-based decision system that accurately diagnoses driving situations from a few vehicle parameters and from traffic light color, to select instantly the appropriate longitudinal action in a rational and explainable way.

Goals & benefits

Identify the car parameters involved in the driving situation diagnosis and rediscover –or confirm– the causal relationships between specific parameters, their combination, and the requested action.

Help engineers design reliable intelligible autonomous vehicles that assist the driver efficiently. Intelligible means that the internal decision logic of the decision system is explicit; therefore this ADAS could be audited by the domain expert and certified by the regulator.

Enforce the use of stable and transparent models before embedding them in the vehicle.

  • The top-model is a decision system composed of 8 gradual rules without chaining, aggregated into 5 disjonctive fuzzy rules.
  • Each rule uses from 1 to 3 predictors among the 4 variables that XTRACTIS identified as significant (out of the 5 potential predictors characterizing driving situations).
  • Only a few rules are triggered at a time to compute the decision.

It has a perfect Real Performance (on unknown data).

It computes real-time predictions up to 70,000 decisions/second, offline or online (API).

UC21-scores-graph v2

LoR=Logistic Regression
RFo=Random Forest
BT=Boosted Tree
NN=Neural Network

Detailed results and explanations in full document

Use Case 2024/02 (v3.1)

Powered by XTRACTIS® REVEAL 13.0.45933 (2023/07)


  1. Problem Definition
  2. XTRACTIS-induced Decision System
  3. XTRACTIS Process
  4. Top-Model Induction
  5. Explained Predictions for 5 unkown cases
  6. Top-Models Benchmark
  7. Appendices