Artificial Force Field for Haptic Feedback in UAV Teleoperation

TLDR: Results indicate that the novel AFF more effectively performs the collision avoidance function than potential fields known from literature, and because of its smaller size, the field yields lower repulsive forces, results in less force cancellation effects, and allows for larger UAV velocities.

Abstract: The feedback upon which operators in teleoperation tasks base their control actions differs substantially from the feedback to the driver of a vehicle. On the one hand, there is often a lack of sensory information; on the other hand, there is additional status information presented via the visual channel. Haptic feedback could be used to unload the visual channel and to compensate for the lack of feedback in other modalities. For collision avoidance, haptic feedback could provide repulsive forces via the control inceptor. Haptic feedback allows operators to interpret the repulsive forces as impedance to their control deflections when a potential for collision exists. Haptic information can be generated from an artificial force field (AFF) that maps environment constraints to repulsive forces. This paper describes the design and theoretical evaluation of a novel AFF, i.e., the parametric risk field, for teleoperation of an uninhabited aerial vehicle (UAV). The field allows adjustments of the size, shape, and force gradient by means of parameter settings, which determine the sensitivity of the field. Computer simulations were conducted to evaluate the effectiveness of the field for collision avoidance for various parameter settings. Results indicate that the novel AFF more effectively performs the collision avoidance function than potential fields known from literature. Because of its smaller size, the field yields lower repulsive forces, results in less force cancellation effects, and allows for larger UAV velocities. This indicates less operator control demand and more effective UAV operations, both expected to lead to lower operator workload, while, at the same time, increasing safety.

Figures

Fig. 11. Schematic representation of the closed-loop autonomous system.

Fig. 10. Definitions of the direction of the risk vector. (a) Radial projection. (b) Normal projection.

Fig. 16. Close-up of some trajectories with BRF contour lines (all with radial projection unless other specified). (a) Trajectory B, G = 0.1 s. (b) Trajectory B, G = 1 s. (c) Trajectory D, G = 0.1 s. (d) Trajectory D, G = 1 s. (e) Trajectory E, G = 1 s. (f) Trajectory E, G = 1 s, normal projection. (g) Trajectory F, G = 0.1 s, normal projection. (h) Trajectory F, G = 1 s, normal projection.

Fig. 1. GPF requires both the distance d to an obstacle and the velocity component vi (of the velocity vector v) toward this obstacle.

Fig. 2. Definition of tmin, tmax, tres, and dstop adopted from [14].

Fig. 15. Simulation results with the BRF for three values of G (1, 0.1, and 0.07 s). The lines represent the traveled path of the UAV for various parameter settings. (a) BRF, radial risk vectors. (b) BRF, normal risk vectors.

Frequently asked questions

Q1. What contributions have the authors mentioned in the paper "Artificial force field for haptic feedback in uav teleoperation" ?

This paper describes the design and theoretical evaluation of a novel AFF, i. e., the parametric risk field, for teleoperation of an uninhabited aerial vehicle ( UAV ). This indicates less operator control demand and more effective UAV operations, both expected to lead to lower operator workload, while, at the same time, increasing safety. Results indicate that the novel AFF more effectively performs the collision avoidance function than potential fields known from literature. 

Q2. What are the future works mentioned in the paper "Artificial force field for haptic feedback in uav teleoperation" ?

Future research will involve the design and tuning of the haptic interface, including human-in-the-loop experiments to evaluate its effects on operator workload and situation awareness. 

Q3. What is the effect of the inclusion of relative velocity on the potential field?

The inclusion of relative velocity prevents the potential field from generating repulsive forces when the vehicle is near an obstacle but, at the same time, not moving toward it. 

Q4. What is the resulting total risk vector?

The resulting total risk vector is represented by the solid arrow in the center of the UAV, whereas the velocity vector is represented by the dashed-dotted arrow in the center of the UAV. 

Q5. What is the effect of the PRF on the risk vector?

It can also be seen that, with normal risk vectors, the PRF resulted in high fluctuations in the risk at the corner of the obstacles, whereas the BRF resulted in two rather large peak values. 

Q6. What was the effect of the large field on the UAV?

With the small field [Fig. 16(c)], the UAV was mainly subjected to the repulsive forces from the dead-end wall, resulting in a more effective deceleration of the UAV. 

Q7. What was the effect of the normal projection on the risk vectors?

For both risk fields, the normal risk vectors resulted in higher velocities in the narrow corridor than with radial projection [see Fig. 14(b)]. 

Q8. What is the reason why the collision in trajectory D would not happen in real life?

the collision in trajectory D would probably not happen in real life and is perhaps an artifact of using a rather simple autonomous control system. 

Q9. What would be the risk vector magnitude for a human operator?

In practice, a close distance would result in a large risk vector magnitude, and a sudden change in the vector direction and magnitude would be noticeable for a human operator. 

Q10. What is the role of haptic feedback in collision avoidance?

In this respect, haptic feedback plays an important role in complementing visual feedback particularly with trajectories A, B, and F. 

Q11. What is the avoidance maneuver of the vehicle?

the avoidance maneuver of the vehicle would mainly depend on the relative position, size, and density of the obstacles and less on the obstacles’ shape. 

Q12. What is the way to present the risk to a human operator?

Aside from the most obvious implementation mentioned previously—presenting the risk through a force offset on the stick—an alternative would be by means of an increase in the spring constant of the manipulator dynamics, i.e., changing the stiffness of the manipulator as a function of the risk [28], [40], [42]. 

Q13. What was the shortest path for the UAV?

2) Trajectory B: Here, the UAV had to move through a corridor that initially had a width of 4rpz, which decreased to 2.7rpz, resulting in a steplike disturbance. 

Q14. How can the authors get the final avoidance force vector?

According to the theory of the GPF, the final avoidance force vector can be obtained by taking the gradient of the potential field for every obstacle inside the field.