OASIS: Optimization-Aware Subgoal Inference with Safety Constraints for Autonomous Mobile Robot Navigation
Authors: Arian Hajizadeh, Javad Pourrafi Arabani, Negar Shenavar, Sana Bashiri, Seyed Majid Esmailzadeh
Organizations: Electrical Engineering, Iran University of Science and Technology, Tehran, Iran · School of Electrical Engineering, Iran University of Science and Technology, Tehran, Iran
Navigation in dynamic environments requires mobile robots to make progress toward their goals while respecting motion limits and avoiding surrounding obstacles. Learned subgoal policies can guide model predictive control beyond its local prediction horizon, but a geometrically suitable subgoal does not necessarily lead to a successfully solved control problem. This paper proposes OASIS, a hierarchical navigation framework that incorporates recent optimization outcomes into reinforcement learning-based subgoal generation. A high-level policy trained using Soft Actor--Critic generates local subgoals, while a nonlinear model predictive controller tracks them subject to motion limits, geometric collision-avoidance constraints, and embedded high-order control barrier function constraints. The outcomes of controller optimizations over each subgoal interval are summarized and provided to the policy at its next decision, without requiring additional optimization calls. The framework is evaluated in five navigation scenarios with static and moving obstacles, including dense interactions and sensing uncertainty. Comparisons with hierarchical and direct-goal control variants show that OASIS achieves the highest aggregate navigation success. Matched comparisons across three training seeds further show a consistent success advantage over the otherwise identical hierarchy without optimization-status input. The selected-policy comparison also shows fewer primary optimization failures with similar mean controller time.
Figures & tables
Figure 1: Proposed architecture. A high-level policy generates subgoals from navigation observations and previous optimization outcomes. A low-level predictive controller tracks these subgoals subject to motion, collision-avoidance, and HOCBF constraints, and returns optimization feedback for the next policy decision.
Figure 2: Overall hierarchical architecture of OASIS. The upper layer uses a Soft Actor-Critic (SAC) framework to generate normalized subgoals ηj based on environment observations and execution feedback qj . The lower layer transforms ηj into workspace subgoals gj and solves a Nonlinear Model Predictive Control problem with High-Order Control Barrier Functions (NMPC–HOCBF) to compute low-level control commands u=[a,α]T while returning solver outcomes qj+1 to inform future subgoal generation.
Scenario
Static
Moving
Nominal mixed
3
3
Static slalom
6
0
Dense-6
4
6
Dense-7
4
7
Uncertain/noisy
3
4
Table 1: Evaluation scenarios.
Figure 3: All scenarios for the same initial and goal positions
Variant
Reference source
Safety Constraint
Op Feedback
Op Penalty
OASIS
SAC subgoal
Yes
Yes
Yes
A+
SAC subgoal
Yes
No
Yes
A
SAC subgoal
Yes
No
No
B+
SAC subgoal
No
Yes
Yes
B
SAC subgoal
No
No
No
C
Final goal
Yes
—
—
Table 2: Comparison groups.
Parameter
Symbol
Value
Sampling interval
Δt
0.1 s
Prediction horizon
N
15 steps
Subgoal execution interval
K
Up to 4 steps
Robot radius
rR
0.25 m
Maximum subgoal distance
dmax
1.35 m
Linear speed bounds
v
[0,1.05] m/s
Table 3: MPC and HOCBF configuration. HOCBF settings apply only to HOCBF-equipped variants.
Method
Success (%)
Collisions (%)
Timeout (%)
Path length (m)
Travel time (s)
QP failure (%)
Controller time (ms)
OASIS
77.40
19.20
3.40
7.84
9.99
3.94
36.26
A+
70.60
25.00
4.40
7.59
9.69
4.51
36.32
A
72.80
23.20
4.00
7.48
9.29
4.52
34.19
B+
68.80
30.80
0.40
7.69
9.16
2.75
4.69
B
59.60
39.60
0.80
8.37
11.09
3.14
6.84
C
68.40
31.20
0.40
6.15
6.70
7.58
38.40
Table 4: Navigation performance across five scenarios. Boundary violations are counted as collisions. Path length and travel time are averaged over successful episodes; Controller time is averaged over control calls.
Method
Nominal mixed
Static slalom
Dense-6
Dense-7
Noisy/uncertain
Overall
OASIS
96
98
60
56
77
77.4
A+
88
100
52
39
74
70.6
A
94
100
61
37
72
72.8
B+
79
100
60
54
51
68.8
B
78
86
48
32
54
59.6
C
82
100
56
43
61
68.4
Table 5: Success rates (%) over 100 episodes per scenario.
Seed
OASIS
A+
A
42
75.80
70.60
71.20
123
77.40
69.20
72.80
456
76.20
69.80
70.80
Mean across three seeds
76.47
69.87
71.60
Table 6: Success rates (%) across training seeds over five scenarios
Figure 4: Representative successful navigation episodes using OASIS in three evaluation scenarios.
Real-time autonomous navigation in dynamic, unknown environments remains a fundamental challenge for mobile robotics. We propose a map-free framework that tightly integrates reactive rolling-horizon planning with nonlinear Model Predictive Control (MPC). At each control cycle, a LiDAR-based Gaussian occupancy representation is constructed and used to generate collision-free trajectories via A* search, which are then tracked by a CasADi/IPOPT MPC formulation incorporating a smooth sigmoid obstacle barrier. To improve robustness to parameter sensitivity, we adopt an offline Bayesian optimization scheme based on Tree-structured Parzen Estimators (TPE), which identifies near-optimal controller parameters with respect to a composite navigation objective. In addition, a Gaussian Process surrogate is used to analyze parameter sensitivity and provide insight into the optimization landscape. The proposed framework is robot-agnostic and is evaluated on the Unitree Go2 quadruped in simulation using Gazebo, followed by deployment on the physical robot. Experimental results show that parameters tuned in simulation transfer effectively to hardware, maintaining comparable performance without additional tuning. The full system achieves up to a 90.0% navigation success rate when deployed, along with a 38.9% average improvement in the evaluation metrics across simulated environments.
Lorenzo Ortolani, Gabriel Voss, Gabriele Beltrami +2
Achieving safe navigation in confined dynamic environments remains a challenging problem, as robots must balance navigation efficiency with safety. Existing actor--critic MPC architectures typically employ differentiable MPC layers, but this way cannot readily accommodate explicit state constraints, making it intractable to involve collision avoidance within learning-based planning. To address this issue, we propose an actor--critic model predictive control framework with state constraints. We first use vertices to represent the robot and obstacles, and half-spaces to represent the workspace. Based on these, the nonlinear robot dynamics is locally linearized and the hyperplane-based collision avoidance constraints are convexified into the differentiable MPC layer, and then the actor exploits closed-loop feedback to adjust the MPC objective parameters that shape local motion behavior, while the critic evaluates long-term task performance. We further relax the convexified constraints with bounded and penalized slack variables to form an elastic quadratic program, accommodating cases where the local constraint approximations cannot be satisfied simultaneously. Simulation tests and experiments, involving dynamic collision avoidance for multi-robot in narrow cases, illustrate the advantages of the proposed framework in adaptively adjusting the learned policy and improving navigation efficiency over existing approaches while maintaining the motion safety.
Chaoran Wu, Jiayu Fan, Qichen Tan +2
School of Future Science and Engineering, Soochow University, Suzhou 215222, China · Chinese Academy of Sciences Shanghai Institute of Microsystem and Information Technology, Shanghai 200233, China
Visual navigation is a core capability for mobile robots, yet end-to-end learning-based methods often struggle with generalization and safety in unseen, cluttered, or narrow environments. These limitations are especially pronounced in dense indoor settings, where collisions are likely and end-to-end models frequently fail. To address this, we propose SaferPath, a hierarchical visual navigation framework that leverages learned guidance from existing end-to-end models and refines it through a safety-constrained optimization-control module. SaferPath transforms visual observations into a traversable-area map and refines guidance trajectories using Model Predictive Stein Variational Evolution Strategy (MP-SVES), efficiently generating safe trajectories in only a few iterations. The refined trajectories are tracked by an MPC controller, ensuring robust navigation in complex environments. Extensive experiments in scenarios with unseen obstacles, dense unstructured spaces, and narrow corridors demonstrate that SaferPath consistently improves success rates and reduces collisions, outperforming representative baselines such as ViNT and NoMaD, and enabling safe navigation in challenging real-world settings.
Lingjie Zhang, Zeyu Jiang, Changhao Chen
PEAK-Lab, The Hong Kong University of Science and Technology (Guangzhou), Guangzhou, 511453, China