1 of 12

RMP Navigation for BARN

Precious Philip-Ifabiyi

Robotics Student Fellow @ Autonomous Systems Lab

​

Supervised by:

Dr. Lionel Ott

Dr. Michael Pantic

�

2 of 12

Table of Contents

  1. Key slides from the first presentation
  2. Benchmarking Reactive Nagivation using RMP
  3. Benchmarking Reactive Navigation using RMP and VFH+
  4. State Estimation for the ESA SUPPORTER project
  5. Results
  6. Conclusion
  7. Future works

13.03.2025

2

3 of 12

What I worked on

13.03.2025

3

Reactive LiDAR navigation¹

Reactive LiDAR navigation using RMP and VFH+2

State Estimation of the SUPPORTER robot

1. M. Pantic, I. Meijer et al.: Obstacle avoidance using raycasting and Riemannian Motion Policies at kHz rates for MAVs, ICRA 2023

Objective: Benchmark and improve RMP navigation algorithms using the BARN dataset.

2. I. Ulrich and J. Borenstein, "VFH+: reliable obstacle avoidance for fast mobile robots," Proceedings. 1998 IEEE International Conference on Robotics and Automation (Cat. No.98CH36146), Leuven, Belgium, 1998, pp. 1572-1577 vol.2, doi: 10.1109/ROBOT.1998.677362.

4 of 12

What I worked on

13.03.2025

4

(i) Reactive LiDAR navigation1

(ii) Reactive LiDAR navigation using RMP and VFH+2

(iii) Comparison

1. M. Pantic, I. Meijer et al.: Obstacle avoidance using raycasting and Riemannian Motion Policies at kHz rates for MAVs, ICRA 2023.

Objective: Benchmark and improve RMP navigation algorithms using the BARN dataset.

2. I. Ulrich and J. Borenstein, "VFH+: reliable obstacle avoidance for fast mobile robots," Proceedings. 1998 IEEE International Conference on Robotics and Automation (Cat. No.98CH36146), Leuven, Belgium, 1998, pp. 1572-1577 vol.2, doi: 10.1109/ROBOT.1998.677362.

5 of 12

Reactive LiDAR navigation

  • Originally developed for OMAVs.
  • Based on Riemannian Motion Policies.
  • Second-order dynamical system governing robot acceleration.
  • Integrates goal-seeking with obstacle avoidance in navigation.

​

13.03.2025

5

6 of 12

Reactive LiDAR navigation using RMP and VFH+

13.03.2025

6

Polar Histogram used by VFH+

7 of 12

What I worked on

13.03.2025

7

Reactive LiDAR navigation using RMP and VFH+

Obstacle Avoidance Policy

Total Policy

Goal Policy

Goal Point

LiDAR scan

Robot

8 of 12

Benchmarking Results

​

13.03.2025

8

The algorithms were evaluated across 50 distinct BARN environments, with 10 runs conducted in each setting.

9 of 12

Benchmarking Results for Reactive Nagivation using only RMP

​

13.03.2025

9

Failures due to:

  • Local Minima

x2.5

10 of 12

Benchmarking Results for Reactive Nagivation using only RMP

​

13.03.2025

10

Failures due to:

  • Nonholonomic constraint

x2.5

11 of 12

Benchmarking Results for Reactive Nagivation using VFH + RMP

​

13.03.2025

11

Local minima problems.

12 of 12

Conclusion

  • Improvements
    • Incorporate the kinematic model of the robot into the RMP.
    • Model the VFH+ algorithm as an RMP.
  • Takeaway
    • Parameter tuning is crucial for optimal performance.

​

13.03.2025

12

RMP (3x)

RMP+VFH (3x)