[OVPHYSX] PVA sensor

[AntoineRichard]

Summary

Implement `Pva` and `PvaData` for the OVPhysX backend, satisfying the `BasePva` and `BasePvaData` contracts. Superset of IMU — provides world-frame pose, body-frame velocities, body-frame accelerations (linear + angular), and projected gravity.

Validation: Standalone script similar to IMU.

Scope

• `pva.py` (~350 lines), `pva_data.py` (~120 lines), `kernels.py` (~150 lines)
• Copy PhysX tests with config swap

No OVPhysX API Blockers

Same data sources as IMU. No new ovphysx APIs needed.

Dependencies

• Requires PR ovphysx backend integration #4852 merged
• Implement after [OVPHYSX] IMU sensor #5318 (IMU) — shares patterns and kernels

Spec

Full design spec: `docs/superpowers/specs/2026-04-20-ovphysx-pva-sensor-design.md`

Parent issue: #5315

---

📋 Full Design Spec (click to expand)

OVPhysX PVA Sensor — Design Spec

Issue: #5319 — [OVPHYSX] PVA sensor
Date: 2026-04-20
Status: Draft

Summary

Implement Pva and PvaData for the OVPhysX backend, satisfying the BasePva and BasePvaData contracts. PVA is a superset of IMU — it provides world-frame pose, body-frame velocities, body-frame linear and angular accelerations, and projected gravity. Uses numerical differentiation for accelerations.

Validation environment: No existing task uses PvaCfg. Write a standalone validation script similar to IMU.

Depends on: Shares patterns with #5318 (IMU) — implement IMU first, then PVA extends it.

Contract to Satisfy

BasePva (source/isaaclab/isaaclab/sensors/pva/base_pva.py)

Inherits SensorBase. Additional abstract:

Property/Method	Type
data	BasePvaData
_initialize_impl()	method
_update_buffers_impl(env_mask)	method

BasePvaData (source/isaaclab/isaaclab/sensors/pva/base_pva_data.py)

Property	Shape	Type	Description
pose_w	(N,)	wp.transformf	World-frame pose
pos_w	(N,)	wp.vec3f	World position [m]
quat_w	(N,)	wp.quatf	World orientation
lin_vel_b	(N,)	wp.vec3f	Body-frame linear velocity [m/s]
ang_vel_b	(N,)	wp.vec3f	Body-frame angular velocity [rad/s]
lin_acc_b	(N,)	wp.vec3f	Body-frame linear acceleration [m/s^2]
ang_acc_b	(N,)	wp.vec3f	Body-frame angular acceleration [rad/s^2]
projected_gravity_b	(N,)	wp.vec3f	Gravity in body frame

Architecture

File Layout

source/isaaclab_ovphysx/isaaclab_ovphysx/sensors/
└── pva/
    ├── __init__.py
    ├── __init__.pyi
    ├── pva.py           (~350 lines)
    ├── pva_data.py      (~120 lines)
    └── kernels.py       (~150 lines)

Implementation Pattern

Identical to IMU but with more outputs:

pva.py:

1. _initialize_impl(): Same as IMU — find ancestor, resolve offset, get gravity, allocate buffers

   • Additional: allocate prev_ang_vel_w for angular acceleration differentiation

2. _update_buffers_impl(env_mask):

   • Read body transforms and velocities from ovphysx bindings
   • Launch pva_update_kernel per environment:
     • Apply sensor offset to get world-frame pose (pos_w, quat_w)
     • COM correction on linear velocity
     • Numerically differentiate both linear and angular velocities
     • Rotate all to body frame
     • Project gravity to body frame: projected_gravity_b = quat_apply_inverse(quat, gravity_vec)
     • Store 7 outputs + update velocity history

pva_data.py:

• Container with 8 warp array properties
• Simple property accessors

OVPhysX API Requirements

No new ovphysx APIs needed — same data sources as IMU (body transforms, velocities, COM).

Warp Kernels

Kernel	Purpose
pva_update_kernel	Per-env: all IMU computation + angular acceleration + pose output + gravity projection
pva_reset_kernel	Zero all 8 output arrays + velocity history

Tests

Source: source/isaaclab_physx/test/sensors/test_pva.py
Target: source/isaaclab_ovphysx/test/sensors/test_pva.py

Validation script: check_pva.py — verify all 8 outputs for a body with known motion.

Dependencies

• Requires PR ovphysx backend integration #4852 merged
• No new ovphysx APIs needed
• Implement after [OVPHYSX] IMU sensor #5318 (IMU) — shares patterns and kernels

Estimated Scope

• pva.py: ~350 lines
• pva_data.py: ~120 lines
• kernels.py: ~150 lines
• Tests: ~150 lines of adaptations