Fall Test Plan

Fall Validation Experiment

Oliver Krengel, Rohit Murthy, Chien Chi Ho, Pengsheng Guo, Vivek G R

Location: NSH B-Level, “east” points outward from elevator hallway, this is the +x axis

Equipment: Sensor Infrastructure, Origin Chronos Desktop PC, Team Lo-Co vehicle, human agent wearing non-black clothing

Step Description Requirement
0.1 Place sensor infrastructure 8 meters south and 4 meters east of SE corner of cage.  Set height to minimum, angle at 0 degrees to horizontal, 45 degrees north of west.  This point is defined as the origin. none
1.1 Measure points (-6,6), (-12,0), (-2,10), (-6,-10), (-10,10), and (0,20) from origin.  Mark these points on the floor with blue tape. Detection error
1.2.1 Human agent stands at floor marker #1.  Display centroid on desktop – measure error as distance from point (8.5,0). Detection error
1.2.2-1.2.5 Repeat step [1.2.1] for points #2-#5, respectively.  Error will be measured from points (8.5,8.5), (8.5,-5.7), (-2.8,11.3), (0,14.1) in the infrastructure frame. Detection error
Test 1 Measure error from ground truth at each point.  

  • Test passes if average error is less than 0.3 meters.
 Detection error
2.1 Human agent stands at floor marker #6 then walks between (0,20) and (-4,20) repeatedly 5 times. Tracking range
Test 2 Display recorded pedestrian trajectory on desktop.

  • Test fails if no centroid is apparent on the desktop.
  • Test fails if centroid extrema are less than 3.4m apart for 3 or fewer of the cycles.
 Tracking range
3.1 Human agent walks continuously in test area.  Capture fifty trajectory predictions and measured against detection 2 seconds in the future. Trajectory prediction
Test 3 Measure estimated position 2 seconds in the future against detected position.

  • Test passes if average error for 5 tests is less than 0.5 meters.
 Trajectory prediction
4.1 Human agent walks out of range to the north, behind wall to restroom if necessary. Cycle time
4.2 Human agent walks back into range toward the infrastructure. Cycle time
Test 4 Display tracking information on desktop.

  • Test passes if first timestamped centroid and first published trajectory are less than 1 second apart.
 Cycle time
5.1 Set vehicle center 1.5 meters east of east cage wall, 18 meters north of infrastructure.  Human agent stands 10 meters west of infrastructure, 1 meter south of south cage wall. Stop short
Test 5 Refer to diagram below for clarity on positioning.

Set vehicle waypoint to (-2.5, 0).  Start vehicle.  Vehicle runs at 3 m/s.

Reset vehicle. Set vehicle waypoint to (-2.5,0).  Start vehicle.  Human agent starts walking east 2 seconds after vehicle start at a speed of ~1.5 m/s.

  • Test passes if vehicle reaches waypoint on first pass, stops short on second pass.
 Stop short

 

Spring Validation Experiment

Oliver Krengel, Rohit Murthy, Chien Chi Ho, Pengsheng Guo, Vivek G R

Requirement
Test
Detect Multiple Pedestrians’ CentroidDisplay 4 pedestrians centroids @ measured floor points
Track Multiple Pedestrians within 20mDisplay 4 pedestrian centroid @ location 20m away
Predict Multiple Trajectories 2 seconds into futureDisplay 4 pedestrian trajectories information while in infrastructure field of view
Cycle time < 0.5 secondsTimestamp first centroid, timestamp time of publish, display difference as cycle time
Plan around multiple pedestriansTest course 1 time without pedestrians
Test 5 times with different multiple pedestrian trajectories