Fix trajectory evaluation

Remove unused imports

src/curobo/wrap/reacher/evaluator.py

@@ -218,12 +218,13 @@ def compute_smoothness(
     scale_dt = (1 / dt_score).view(-1, 1, 1)
     abs_acc = torch.abs(acc) * (scale_dt**2)
     # mean_acc_val = torch.max(torch.mean(abs_acc, dim=-1), dim=-1)[0]
-    max_acc_val = torch.max(torch.max(abs_acc, dim=-1)[0], dim=-1)[0]
+    max_acc_val = torch.max(abs_acc, dim=-2)[0] # batch x dof
     abs_jerk = torch.abs(jerk) * scale_dt**3
     # calculate max mean jerk:
     # mean_jerk_val = torch.max(torch.mean(abs_jerk, dim=-1), dim=-1)[0]
-    max_jerk_val = torch.max(torch.max(abs_jerk, dim=-1)[0], dim=-1)[0]
+    max_jerk_val = torch.max(abs_jerk, dim=-2)[0]  # batch x dof
     acc_label = torch.logical_and(max_acc_val <= max_acc, max_jerk_val <= max_jerk)
+    acc_label = torch.all(acc_label, dim=-1)
     return (acc_label, smooth_cost(abs_acc, abs_jerk, dt_score))
 
 

tests/motion_gen_eval_test.py

@@ -0,0 +1,60 @@
+#
+# Copyright (c) 2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# NVIDIA CORPORATION, its affiliates and licensors retain all intellectual
+# property and proprietary rights in and to this material, related
+# documentation and any modifications thereto. Any use, reproduction,
+# disclosure or distribution of this material and related documentation
+# without an express license agreement from NVIDIA CORPORATION or
+# its affiliates is strictly prohibited.
+#
+
+# Third Party
+import pytest
+import torch
+
+# CuRobo
+from curobo.types.base import TensorDeviceType
+from curobo.types.robot import JointState
+from curobo.wrap.reacher.motion_gen import MotionGen, MotionGenConfig, MotionGenPlanConfig
+from curobo.wrap.reacher.evaluator import TrajEvaluatorConfig
+
+@pytest.fixture(scope="module", params=[True, False])
+def evaluate_interpolated_trajectory(request):
+    return request.param
+
+@pytest.fixture(scope="module")
+def motion_gen(evaluate_interpolated_trajectory):
+    tensor_args = TensorDeviceType()
+    world_file = "collision_test.yml"
+    robot_file = "franka.yml"
+    dof = 9
+    traj_evaluator_config = TrajEvaluatorConfig(
+         max_acc=torch.ones((dof), device=tensor_args.device, dtype=tensor_args.dtype),
+         max_jerk=torch.ones((dof), device=tensor_args.device, dtype=tensor_args.dtype),
+         min_dt=torch.tensor(0.01, device=tensor_args.device, dtype=tensor_args.dtype),
+         max_dt=torch.tensor(1.5, device=tensor_args.device, dtype=tensor_args.dtype)
+    )
+    motion_gen_config = MotionGenConfig.load_from_robot_config(
+        robot_file,
+        world_file,
+        tensor_args,
+        trajopt_tsteps=26,
+        use_cuda_graph=False,
+        num_trajopt_seeds=50,
+        fixed_iters_trajopt=True,
+        evaluate_interpolated_trajectory=evaluate_interpolated_trajectory,
+        traj_evaluator_config=traj_evaluator_config,
+    )
+    motion_gen = MotionGen(motion_gen_config)
+    motion_gen.warmup(warmup_js_trajopt=True)
+
+    retract_cfg = motion_gen.get_retract_config()
+
+    start_state = JointState.from_position(retract_cfg.view(1, -1).clone())
+    goal_state = JointState.from_position(retract_cfg.view(1, -1).clone())
+
+    result = motion_gen.plan_single_js(start_state, goal_state, MotionGenPlanConfig(max_attempts=1))
+
+def test_motion_gen(motion_gen):
+    return True