Semantic obstacle recognition for path planning
Systems, methods, devices, and techniques for planning travel of an autonomous robot. A system identifies one or more obstacles that are located in proximity of at least a portion of a planned route for the autonomous robot. For each obstacle, the system: (i) determines a semantic class of the obstacle, including selecting the semantic class from a library that defines a set of multiple possible semantic classes for obstacles, and (ii) selects a planning policy for the obstacle that corresponds to the semantic class of the obstacle. The system can generate a trajectory along the at least the portion of the planned route using the selected planning policies. The robot can then initiate travel according to the trajectory.
1. A computer-implemented method, comprising:
identifying a particular obstacle in a proximity of an autonomous robot;
determining a semantic class of the particular obstacle, including selecting the semantic class from a plurality of pre-defined semantic classes;
identifying an obstacle cost function that corresponds to the semantic class of the particular obstacle, wherein the obstacle cost function defines a cost of traveling relative to a given obstacle within the semantic class of the particular obstacle, wherein the cost varies based at least on (i) a position of the autonomous robot relative to the given obstacle and (ii) a speed of the autonomous robot; and
directing the autonomous robot to travel relative to the particular obstacle using the cost defined by the obstacle cost function.
2. The computer-implemented method of claim 1 , comprising using the obstacle cost function to plan a trajectory for the autonomous robot to travel relative to the particular obstacle, wherein directing the autonomous robot to travel relative to the particular obstacle comprises directing the robot to initiate travel according to the planned trajectory.
3. The computer-implemented method of claim 2 , wherein planning the trajectory for the autonomous robot comprises optimizing a cost of the trajectory, and the cost of the trajectory is a function of the cost defined by the obstacle cost function.
4. The computer-implemented method of claim 1 , wherein the obstacle cost function defines different values for the cost of traveling relative to the given obstacle for at least two different positions of the autonomous robot relative to the given obstacle when the autonomous robot is traveling at a same speed at the at least two different positions.
5. The computer-implemented method of claim 1 , wherein the obstacle cost function defines different values for the cost of traveling relative to the given obstacle for at least two different speeds of the autonomous robot when the autonomous robot is located at a same position relative to the given obstacle.
6. The computer-implemented method of claim 1 , wherein varying the cost defined by the obstacle cost function based on the position of the autonomous robot relative to the given obstacle comprises varying the cost based on a distance of the autonomous robot from the given obstacle.
7. The computer-implemented method of claim 1 , wherein the obstacle cost function is described in a planning policy associated with the semantic class of the particular obstacle.
8. The computer-implemented method of claim 1 , wherein different ones of the plurality of pre-defined semantic classes are associated with different obstacle cost functions, and identifying the obstacle cost function that corresponds to the semantic class of the particular obstacle comprises selecting an obstacle cost function associated with the semantic class of the particular obstacle.
9. The computer-implemented method of claim 1 , wherein determining the semantic class of the particular obstacle comprises:
accessing a hierarchical representation of the plurality of pre-defined semantic classes that organizes semantic classes based on levels of specificity; and
locating a most-specific semantic class assigned to the particular obstacle.
10. The computer-implemented method of claim 1 , wherein the particular obstacle is a person or an inanimate object.
11. A system comprising:
one or more processors; and
one or more computer-readable media having instructions stored thereon that, when executed by the one or more processors, cause the one or more processors to perform operations comprising:
identifying a particular obstacle in a proximity of an autonomous robot;
determining a semantic class of the particular obstacle, including selecting the semantic class from a plurality of pre-defined semantic classes;
identifying an obstacle cost function that corresponds to the semantic class of the particular obstacle, wherein the obstacle cost function defines a cost of traveling relative to a given obstacle within the semantic class of the particular obstacle, wherein the cost varies based at least on (i) a position of the autonomous robot relative to the given obstacle and (ii) a speed of the autonomous robot; and
directing the autonomous robot to travel relative to the particular obstacle using the cost defined by the obstacle cost function.
12. The system of claim 11 , wherein the operations comprise using the obstacle cost function to plan a trajectory for the autonomous robot to travel relative to the particular obstacle, wherein directing the autonomous robot to travel relative to the particular obstacle comprises directing the robot to initiate travel according to the planned trajectory.
13. The system of claim 12 , wherein planning the trajectory for the autonomous robot comprises optimizing a cost of the trajectory, and the cost of the trajectory is a function of the cost defined by the obstacle cost function.
14. The system of claim 11 , wherein the obstacle cost function defines different values for the cost of traveling relative to the given obstacle for at least two different positions of the autonomous robot relative to the given obstacle when the autonomous robot is traveling at a same speed at the at least two different positions.
15. The system of claim 11 , wherein the obstacle cost function defines different values for the cost of traveling relative to the given obstacle for at least two different speeds of the autonomous robot when the autonomous robot is located at a same position relative to the given obstacle.
16. The system of claim 11 , wherein varying the cost defined by the obstacle cost function based on the position of the autonomous robot relative to the given obstacle comprises varying the cost based on a distance of the autonomous robot from the given obstacle.
17. The system of claim 11 , wherein the obstacle cost function is described in a planning policy associated with the semantic class of the particular obstacle.
18. The system of claim 11 , wherein different ones of the plurality of pre-defined semantic classes are associated with different obstacle cost functions, and identifying the obstacle cost function that corresponds to the semantic class of the particular obstacle comprises selecting an obstacle cost function associated with the semantic class of the particular obstacle.
19. The system of claim 11 , wherein determining the semantic class of the particular obstacle comprises:
accessing a hierarchical representation of the plurality of pre-defined semantic classes that organizes semantic classes based on levels of specificity; and
locating a most-specific semantic class assigned to the particular obstacle.
20. One or more non-transitory computer-readable media having instructions stored thereon that, when executed by the one or more processors, cause the one or more processors to perform operations comprising:
identifying a particular obstacle in a proximity of an autonomous robot;
determining a semantic class of the particular obstacle, including selecting the semantic class from a plurality of pre-defined semantic classes;
identifying an obstacle cost function that corresponds to the semantic class of the particular obstacle, wherein the obstacle cost function defines a cost of traveling relative to a given obstacle within the semantic class of the particular obstacle, wherein the cost varies based at least on (i) a position of the autonomous robot relative to the given obstacle and (ii) a speed of the autonomous robot; and
directing the autonomous robot to travel relative to the particular obstacle using the cost defined by the obstacle cost function.