9.3. Polynomial Via Point Trajectories

This video introduces robot trajectories passing through via points based on cubic polynomial interpolation.

In the last two videos we learned how to define straight-line paths and then time scale them to get trajectories. If we want more flexibility to design the shape of the path, as well as the speed with which it is executed, we could specify a set of configurations through which we would like the robot to transit. These configurations are called via points. We also specify the times at which the robot should achieve each of these via points. We then solve for a smooth trajectory that passes through the via points at the specified times. The choice of the via points and times allows us to shape the path and trajectory. In this case, we solve directly for a trajectory; we do not first find a path and then time scale it.

Let's consider motion in an n-dimensional joint space. For joint i, moving between via points j and j-plus-one, we could define the motion as a third-order polynomial of time. We then apply four terminal constraints, the initial and final position and the initial and final velocity of joint i, to solve for the four coefficients of the polynomial. This is third-order polynomial interpolation with specified via times and velocities. This figure shows a path designed for a two-joint robot using four via points: the start point, the end point, and two other vias. Each via point has the time that the robot passes through the configuration as well as the velocity at that time. The velocity is indicated by the dashed arrows. Each segment between via points, for each degree of freedom, has four coefficients and four terminal constraints, which allows us to solve exactly for the trajectories between via points. The tangent of the path has to be aligned with the specified velocity at each via point, so we can use the velocities at the via points to change the shape of the path. These time plots show the position and velocity of each joint during the trajectory. You can see that the positions and velocities are continuous at the via points, but the acceleration is not.

Discontinuity in the acceleration may not be desirable. Also, it may be cumbersome to have to specify the velocity at each via point. Therefore, another solution is to leave the velocities at the via points free, but to constrain the velocity before and after a via point, and the acceleration before and after a via point, to be equal. This is third-order polynomial interpolation with specified via times only.

There are other ways to shape a path using via points or control points. In particular, with B-splines, the path does not pass exactly through the control points, but the path is guaranteed to remain in the convex hull of the control points, unlike the path you see here. This property ensures that the path does not violate joint limits.