Wrong time step width in plots generated by Problem.plot_trajectories()

[chris-konrad]

The time array for trajectory plotting created by Problem.plot_trajectories() covers the interval [0, num_collocation_nodes * node_time_interval] including the limits while forcing the array size to be num_collocation_nodes. As a result, the time spacing is not as expected. This is the default behaviour of np.linspace(). Should be fixed by using np.arange() instead.

Example for num_collocation_nodes = 5, node_time_interval=1:

In[2]: np.linspace(0, num_collocation_nodes * node_time_interval, num=num_collocation_nodes)
Out[2]: array([0.  , 1.25, 2.5 , 3.75, 5.  ])

In[3]: np.arange(0, num_collocation_nodes * node_time_interval, node_time_interval)
Out[3]: array([0, 1, 2, 3, 4])

In the linspace version, the time steps of the resulting array are not 1, which however would be expected because of node_time_interval=1.

[Peter230655]

As I know it, the final time is defined as tf = node_time_interval * (num_collection_nodes - 1),
you defined it as tf = node_time_interval * num_collection_nodes

[chris-konrad]

That would fix the problem as well. But it is currently not implemented that way (see: /opty/direct_collocation.py#L400).

[Peter230655]

Maybe you are right, but all the examples I ever saw, the final time was defined a interval * (num-nodes - 1), assuming the starting time is zero.

[chris-konrad]

Have a look at np.arange(start, stop, step). It defines the interval as [start,stop). In np.arange(0, num_collocation_nodes * node_time_interval), the end of the interval is stop = num_collocation_nodes * node_time_interval = 5. But this will not be included in the time array. It stops at 4:

In[3]: np.arange(0, num_collocation_nodes * node_time_interval, node_time_interval)
Out[3]: array([0, 1, 2, 3, 4])

[chris-konrad]

Solved with #283