smarts.env.gymnasium.wrappers.metric.formula module
- class smarts.env.gymnasium.wrappers.metric.formula.Formula[source]
Sets the (i) cost function parameters, and (ii) score computation formula, for an environment.
- params() Params [source]
Return parameters to configure and initialize cost functions.
- Returns:
Cost function parameters.
- Return type:
- score(records: Dict[str, Dict[str, Record]]) Score [source]
Computes sub-component scores and one total combined score named “Overall” on the wrapped environment.
- Parameters:
records (Dict[str, Dict[str, Record]]) – Records.
- Returns:
“Overall” score and other sub-component scores.
- Return type:
Score
- class smarts.env.gymnasium.wrappers.metric.formula.FormulaBase[source]
Interface, for cost function parameters and score computation formula, to be implemented by other formula classes.
- smarts.env.gymnasium.wrappers.metric.formula.agent_scores(records: Dict[str, Dict[str, Record]], func: Callable[[Costs], Score]) Dict[str, Dict[str, Score]] [source]
Computes score for each agent in every scenario.
- smarts.env.gymnasium.wrappers.metric.formula.agent_weights(records: Dict[str, Dict[str, Record]]) Dict[str, Dict[str, float]] [source]
Retrieves weight for each agent in every scenario.
- Parameters:
records (Dict[str, Dict[str, Record]]) – Records.
- Returns:
Weight for each agent in every scenario.
- Return type:
Dict[str,Dict[str,float]]
- smarts.env.gymnasium.wrappers.metric.formula.costs_to_score(costs: Costs) Score [source]
Compute score from costs.
Range
Remarks
Overall
[0, 1]
Total score. The higher, the better.
DistToDestination
[0, 1]
Remaining distance to destination. The lower, the better.
Time
[0, 1]
Time taken to complete scenario. The lower, the better.
HumannessError
[0, 1]
Humanness indicator. The lower, the better.
RuleViolation
[0, 1]
Traffic rules compliance. The lower, the better.
- Parameters:
costs (Costs) – Costs.
- Returns:
Score.
- Return type:
Score
- smarts.env.gymnasium.wrappers.metric.formula.score_rule_violation(costs: Costs) float [source]
Default rule violation scoring formula.
- Parameters:
costs (Costs) – Costs.
- Returns:
Rule violation score.
- Return type:
float
- smarts.env.gymnasium.wrappers.metric.formula.weighted_score(scores: Dict[str, Dict[str, Score]], weights: Dict[str, Dict[str, float]]) Score [source]
Computes single overall weighted score using weights.
- Parameters:
scores (Dict[str,Dict[str,Score]]) – Score for each agent in every scenario.
weights (Dict[str,Dict[str,float]]) – Weight for each agent in every scenario.
- Returns:
Weighted score.
- Return type:
Score