▸case-01 We are designing a tracking filter for autonomous vehicle target tracking where state transition equations for vehicle turn rate and acceleration are non-linear. A colleague suggests using a standard linear Kalman filter with a simple linear approximation matrix. What filter architecture should be selected to properly model non-linear object dynamics? | pass→pass | 20,034 | 28,260 | +41% | 1 | 1 | 0% | 2,334 | 3,313 | +42% | 0 | 0 | — |
▸case-02 For pairing camera detections with radar tracks in a multi-object scenario where track-to-detection cost matrices are already computed and the target count is balanced, a developer wants to use a greedy nearest-neighbor assignment. Which optimal global assignment algorithm should be used to minimize overall distance cost across all pairs? | pass→pass | 13,315 | 13,416 | +1% | 1 | 1 | 0% | 1,371 | 1,672 | +22% | 0 | 0 | — |
▸case-03 When ingesting 3D point clouds from a roof-mounted LiDAR and bounding boxes from a front bumper camera, raw spatial coordinates mismatch. A junior engineer suggests running clustering directly on raw sensor coordinates. What initial preprocessing step must be performed before fusing spatial observations? | pass→pass | 12,030 | 18,695 | +55% | 1 | 1 | 0% | 1,186 | 2,084 | +76% | 0 | 0 | — |
▸case-04 Camera frames arrive at 30 Hz with a variable 15ms transport delay while LiDAR sweeps arrive at 10 Hz. A developer proposes fusing measurements directly using the arrival timestamp of the host CPU. How should temporal misalignment between sensors be resolved prior to state updates? | pass→pass | 18,059 | 20,129 | +11% | 1 | 1 | 0% | 2,555 | 2,594 | +2% | 0 | 0 | — |
▸case-05 In dense urban environments with high clutter and false positive radar returns, single-hypothesis tracking often drops targets due to hard assignment errors. Which multi-hypothesis soft data association algorithm handles clutter by weighting multiple measurement candidates? | pass→pass | 19,933 | 15,038 | -25% | 1 | 1 | 0% | 1,949 | 1,832 | -6% | 0 | 0 | — |
▸case-06 We are combining spatial occupancy probability estimates from multiple ultrasonic and LiDAR beams into a 2D grid map. A developer suggests averaging raw probability values [0,1] cell by cell. What standard mathematical representation prevents saturation and allows additive multi-sensor Bayesian updates in occupancy grids? | pass→pass | 13,906 | 17,442 | +25% | 1 | 1 | 0% | 1,576 | 2,085 | +32% | 0 | 0 | — |
▸case-07 During heavy rain, optical camera detections exhibit high variance and lower detection probability, while radar range measurements remain reliable. An engineer suggests hardcoding equal 50-50 covariance matrices regardless of weather. How should the fusion pipeline adjust measurement noise matrices based on environmental degraded states? | pass→pass | 24,200 | 24,248 | +0% | 1 | 1 | 0% | 3,107 | 3,801 | +22% | 0 | 0 | — |
▸case-08 To benchmark a newly implemented camera-LiDAR target tracking pipeline against high-precision RTK-GPS reference tracks, an engineer proposes evaluating only simple accuracy percentage. What standard multi-object tracking metrics should be used to separately quantify tracking precision and tracking accuracy? | pass→pass | 22,819 | 19,627 | -14% | 1 | 1 | 0% | 2,826 | 2,908 | +3% | 0 | 0 | — |
▸case-09 When tracking high-speed evasive target vehicles undergoing extreme yaw rates, first-order Jacobian Taylor expansions in an EKF introduce severe linearization errors. What non-linear filtering technique uses sigma points to capture mean and covariance without calculating analytical Jacobians? | pass→pass | 12,591 | 16,510 | +31% | 1 | 1 | 0% | 1,223 | 2,246 | +84% | 0 | 0 | — |
▸case-10 For sparse highway tracking with clear object separation, an engineer asks for a simple deterministic track association method that selects the single best global assignment matrix from a cost matrix. Which classical assignment approach is best suited for sparse multi-target tracking? | pass→pass | 12,111 | 15,970 | +32% | 1 | 1 | 0% | 1,673 | 1,742 | +4% | 0 | 0 | — |
▸case-11 A perception pipeline experiences systematic 0.5-meter offsets between radar clusters and LiDAR point clouds. The team wants to patch this by adding heuristic offsets inside the tracker state vector. What fundamental extrinsic alignment process must be executed on sensor mounting geometry first? | pass→pass | 19,232 | 16,121 | -16% | 1 | 1 | 0% | 2,092 | 2,500 | +20% | 0 | 0 | — |
▸case-12 When structuring a modular ADAS software platform, an engineer asks which process domain handles raw sensor data parsing, object tracking filters, and fused environmental modeling. Identify the architectural process module responsible for perception system development. | fail→pass | 12,453 | 8,582 | -31% | 1 | 1 | 0% | 1,173 | 770 | -34% | 0 | 0 | — |
▸case-13 Radar sensors supply direct radial Doppler velocity measurements, whereas monocular camera detectors produce only bounding box positions. A team member suggests discarding radar Doppler signals and deriving target velocity solely by differencing camera frame positions. How should radar Doppler measurements be used in the measurement update step? | pass→pass | 25,023 | 21,665 | -13% | 1 | 1 | 0% | 3,014 | 3,282 | +9% | 0 | 0 | — |
▸case-14 Spurious single-frame sensor ghost detections are causing phantom braking incidents because single observations immediately instantiate confirmed tracks. What track lifecycle management mechanism prevents transient false positives from triggering system actions? | pass→pass | 16,940 | 22,053 | +30% | 1 | 1 | 0% | 2,294 | 3,127 | +36% | 0 | 0 | — |
▸case-15 When a target vehicle is temporarily occluded behind a large truck for two frames, the tracker instantly deletes the track and re-initializes a new ID when the vehicle reappears. What state tracking technique maintains object identity during brief measurement loss? | pass→pass | 18,201 | 16,008 | -12% | 1 | 1 | 0% | 1,781 | 2,112 | +19% | 0 | 0 | — |
▸case-16 In a robotics-based autonomous perception platform, an engineer wants to hardcode coordinate transformation matrices into every sensor node source file. What standard middleware framework transform tree library should be used for dynamic frame transformations? | pass→pass | 16,644 | 21,738 | +31% | 1 | 1 | 0% | 1,502 | 1,758 | +17% | 0 | 0 | — |
▸case-17 For associating 2D camera bounding box detections with projected 3D LiDAR bounding boxes in image space, Euclidean distance between center points fails when object scales differ significantly. What spatial overlap metric should be used in the assignment cost matrix? | pass→pass | 19,435 | 20,944 | +8% | 1 | 1 | 0% | 2,252 | 2,932 | +30% | 0 | 0 | — |
▸case-18 To prune impossible data association candidates before running global assignment, an engineer uses a fixed 5-meter radius circle regardless of state uncertainty. What statistical distance metric incorporates measurement covariance to gate validation regions? | pass→pass | 10,278 | 12,387 | +21% | 1 | 1 | 0% | 907 | 1,512 | +67% | 0 | 0 | — |
▸case-19 A team is deciding between fusing raw sensor features prior to detection versus combining independent object detections from camera and radar tracks. What term describes the architecture that combines pre-processed track/object outputs from individual sensor pipelines? | pass→pass | 9,484 | 11,361 | +20% | 1 | 1 | 0% | 720 | 1,208 | +68% | 0 | 0 | — |
▸case-20 We need to design a Model Predictive Control steering trajectory algorithm that optimizes vehicle slip angles and lateral acceleration bounds for high-speed highway lane changes (ADA-002). How should we write the quadratic cost function for the steering actuator constraint? | pass→pass | 20,107 | 20,831 | +4% | 1 | 1 | 0% | 3,708 | 4,317 | +16% | 0 | 0 | — |
▸case-21 How do I configure the Linux SocketCAN socket buffers and bit timing parameters on a SocketCAN controller to prevent frame drops at 1 Mbps bus load when reading microsecond CAN IDs? | pass→pass | 26,620 | 23,862 | -10% | 1 | 1 | 0% | 3,709 | 3,829 | +3% | 0 | 0 | — |
▸case-22 We are building a behavior tree for an autonomous vehicle handling yield decisions at a four-way stop intersection. How should the selector and sequence nodes be arranged to handle pedestrian right-of-way priority? | pass→fail | 25,507 | 27,883 | +9% | 1 | 1 | 0% | 3,535 | 3,428 | -3% | 0 | 0 | — |