Benefits to Pedestrian Safety through Intersection Collision Rate and Collision Speed Assessed for Automated Emergency Braking Technology with Open-source Virtual Simulator.
Nationwide, United States
Evaluating the Effects of Cooperative Perception on Avoiding Pedestrian Crashes for Connected and Automated Vehicles - Using Virtual Simulator to Evaluate the Automated Emergency Braking System for Avoiding Pedestrian Crash at Intersections
Summary Information
With Automatic Emergency Braking (AEB) technology, if vehicle sensor detects an imminent collision, the vehicle will break automatically. AEB is expected to benefit pedestrian safety by preventing pedestrian-related crashes caused by human error, but its effectiveness is highly dependent on factors such as the Field of View (FoV), the system software, and the system hardware. This study aimed to:
- explore the effectiveness of AEB in avoiding pedestrian crashes under various occlusion conditions; and
- use CARLA, an open-source virtual simulator, to integrate the AEB control algorithm and sensor for evaluation.
METHODOLOGY
CARLA is an open-source virtual simulator to test automated driving used in this study, featuring a Traffic Manager (TM), command array, Agent Lifecycle and State Management (ALSM), Path Buffer and Vehicle Tracking (PBVT), sensors, and a motion planner stage. The AEB system consisted of two main parts: the sensor model and the braking strategy that used the minimal distance between the vehicle and moving objects to determine the Time-to-Collision (TTC) that would dictate braking decisions. If the TTC was under a designated threshold, the vehicle would decelerate to avoid a collision.
Three scenarios were studied at a typical four-leg intersection:
- Scenario 1: an occluded pedestrian on the near-side crosswalk in the left turn lane
- Scenario 2: an occluded pedestrian on the far-side crosswalk in the left turn lane, and
- Scenario 3: an occluded pedestrian on the near-side crosswalk in the right turn lane.
For the study, 216 motion states that included various initial speed levels of pedestrians and vehicles, different offset times, and 16 AEB control cases (i.e., 15 with AEB control and 1 without AEB control) were used in simulation runs.
Parameter | Value | Step Size | Counts |
---|---|---|---|
Pedestrian initial speed (feet/s) | 2-12 | 2 | 6 |
Ego vehicle initial speed (through, mph) | 25-50 | 5 | 6 |
Offset time for pedestrian to cross (s) | 1-6 | 1 | 6 |
Total | - | - | 216 |
In total, 10,368 (3 scenarios, 216 motion states, 16 AEB control cases) simulation runs were conducted to evaluate the effects of AEB under occlusion conditions.
FINDINGS
Scenario | 1-second TTC | 2-second TTC | 3-second TTC |
---|---|---|---|
Scenario 1 | 15.19% | 30.37% | 32.87% |
Scenario 2 | 9.35% | 14.91% | 18.52% |
Scenario 3 | 31.39% | 41.39% | 45.37% |
Additionally, in collisions that still occurred with AEB usage, the total speed of impact was significantly reduced in all tested scenarios. For example, for a near-side left lane occluded pedestrian, collision speeds were reduced from around 40 miles per hour on average to between 10 and 25 miles per hour on average depending on the allotted reaction time.