Executor-aware Candidate Selection via a Feasibility Certificate

Sooin Choi; Soonwoong Hwang; Wansoo Kim

cs.ROSep 29, 2026

Executor-aware Candidate Selection via a Feasibility Certificate

Authors: Sooin Choi, Soonwoong Hwang, Wansoo Kim

Organizations: Department of Robotics, Hanyang University, Seoul, Republic of Korea · Department of Robotics, Hanyang University ERICA, Ansan, Republic of Korea

Abstract

Modular robotic systems often separate motion planning from a downstream executor that enforces state-dependent hard constraints. A candidate that is geometrically valid may therefore be incompatible with the executor's available command set. We present a certificate-based candidate-selection framework that constructs a command witness from the executor hard set at predicted rollout states and verifies it against the original constraints, without changing candidate generation, ranking, or the executor. Across 5,085 geometry-valid numerical evaluations on two robot models, 795 admitted no executor-feasible command. The certificate is sufficient but conservative: none of the 795 was certified, while 7.09% of reference-feasible cases remained uncertified. In controlled FR3 and fixed-base RB-Y1 simulations, certificate admission frequently changed candidate selection, and a post-hoc exact linear-programming (LP) admission baseline revealed platform-dependent conservatism. Relative to geometry-based selection, certificate admission was associated with lower planner-command coverage and higher nominal tracking error, without a consistent advantage in reached-state interaction reserve. A planner-generated MoveIt/OMPL study further evaluates the same admission rule on externally generated candidate pools.

Figures & tables

Fig. 1: Concept of executor-aware candidate selection. (a) From the same current state $x_{0}$ , the planner generates geometrically valid candidate rollouts toward a common task goal. In the illustrated example, the geometry-based selection $c_{G}^{*}$ contains an executor-infeasible predicted state, whereas certificate-based admission selects $c_{\mathrm{cert}}^{*}$ . Candidate rollouts are visualized by their end-effector paths for clarity; certification is evaluated using the full predicted robot state and executor hard constraints. (b) Under the zero-tolerance idealization, $\mathcal{C}_{\mathrm{cert},0}\subseteq\mathcal{C}_{F}\subseteq\mathcal{C}_{G}$ . In implementation, a constructed witness is accepted after validation against the original hard set within the residual tolerance $\epsilon_{\mathrm{cert}}$ ; candidate generation and ranking remain unchanged.

Fig. 2: Geometry-based and executor-feasibility-aware candidate selection. The proposed framework evaluates the executor hard-constraint formulation at predicted rollout states before selection, while the fixed HQP executor applies the same formulation at the current state.

Robot	Valid	Ref. feas.	Ref. infeas.	Certified	Uncert.
		(%)	(%)	(%)	(%)
Panda	2,673	2,579 (96.48)	94 (3.52)	2,469 (92.37)	204 (7.63)
RB-Y1	2,412	1,711 (70.94)	701 (29.06)	1,517 (62.89)	895 (37.11)
Overall	5,085	4,290 (84.37)	795 (15.63)	3,986 (78.39)	1,099 (21.61)

TABLE I: Reference feasibility and certificate outcomes

Fig. 4: Validation of the feasibility certificate. (a) Cross-classification of certificate decisions and reference feasibility labels. (b) Certificate margin by reference-feasibility class, shown as the dimensionless display quantity of Sec. III-A . The 16 construction failures are annotated separately rather than plotted as finite margins.

	Method	Coverage	Matched gap	Tracking Error	$\rho_{\mathrm{int}}^{5\%}$	Neg. reserve
		[%]	[rad/s]	[mm]	[m/s]	[%]
FR3
	Geom.	100.00	0.2551	14.767	$-0.0200$	30.3
	Current	59.19	0.1920	19.477	$-0.0200$	30.3
RB-Y1
	Geom.	100.00	0.2191	5.360 / 6.468	0.0180	0.37

TABLE II: Interaction-window closed-loop behavior (RB-Y1 EE: left / right)

Explore similar work

Sampling-based Certified Planning with Graphs of Convex Sets
Aug 30, 2026Peng Xie, Amr AlanwarCollision-Free Trajectories Convex Sets
TUM School of Computation, Information and Technology, Department of Computer Engineering, Technical University of Munich, 74076 Heilbronn, Germany
Verified Task-Space Motion Planning Under Joint-Space Constraints
May 21, 2026Hanjiang Hu, Changliu Liu, Yebin WangMotion Planning Low-Latency Latent Planning
Robotics Institute, Carnegie Mellon University · Mitsubishi Electric Research Laboratories (MERL), Cambridge, MA 02139, USA
Execution-Aware Pre-Execution Ranking for Grasp-Conditioned Robotic Placement
Sep 17, 2026Tianyuan Liu, Rutherford Agbeshi Patamia, Benjamin Champion +2Robotic Grasping End-Effector Pose
Deakin University, Australia

Aug 30, 2026, cs.ROJ/K move · Enter open · S save

Sampling-based Certified Planning with Graphs of Convex Sets

Peng Xie, Amr Alanwar

TUM School of Computation, Information and Technology, Department of Computer Engineering, Technical University of Munich, 74076 Heilbronn, Germany

Collision-Free Trajectories Convex Sets Planning

Open paper AskPDF arXiv

Planners on graphs of convex sets return trajectories that are collision-free by construction, provided the convex regions are collision-free. The region generator only promises that property probabilistically, and no planner in the family verifies it. We report the first measurement of what the gap costs. On a scaled 14-DOF bimanual library,

3.2\%

of interface samples are in collision, and a search-based GCS planner (\gcsstar) turns that volume error into a

62\%

answer error:

18

29

pick-and-place queries return trajectories that drive the arms through the shelves, up to

91

,mm deep, reported as successes. Repairing the library does not work; a ten times stricter acceptance contract, sums-of-squares certified regions, and uniform margins each destroy the connectivity planning needs before they deliver soundness. We instead build a planner that certifies its answers. It samples the overlaps and shared faces of the decomposition, prunes with an admissible informed bound, and verifies the one candidate each search round proposes, continuously, by a chain of clearance certificate balls with no resolution parameter; failures are repaired with local in-region detours, and the convex polish is re-verified. Head-to-head on all

29

task queries it delivers zero invalid answers against

21

for the reference, reaches its first certified answer in

0.11

,s against

1.59

,s for the reference's unverified one, and reproduces the reference optimum exactly on every query whose reference answer is physically valid.

Abstract

Explore similar work

Sampling-based Certified Planning with Graphs of Convex Sets

Verified Task-Space Motion Planning Under Joint-Space Constraints

Execution-Aware Pre-Execution Ranking for Grasp-Conditioned Robotic Placement