OBJECTIVES: Artificial Intelligence models are increasingly used in health care, yet global performance metrics can mask variations in reliability across individual patients or subgroups with shared attributes, called patient profiles. This study introduces predictive performance precision analysis in medicine (MED3pa), a method that identifies when models are less reliable, allowing clinicians to better assess model limitations. MATERIALS AND METHODS: We propose a framework that estimates predictive confidence using 3 combined approaches: individualized (IPC), aggregated (APC), and mixed predictive confidence (MPC). Individualized predictive confidence estimates confidence for each patient, APC assesses it across profiles, and MPC combines both. We evaluate our method on 4 datasets: 1 simulated, 2 public, and 1 private clinical dataset. Metrics by declaration rate curves show how performance changes when retaining only the most confident predictions, while interpretable decision trees reveal profiles with higher or lower model confidence. RESULTS: We demonstrate our method in internal, temporal, and external validation settings, as well as through a clinical example. In internal validation, limiting predictions to the 93% most confident cases improved sensitivity by 14.3% and the area under the receiver operating characteristic curve by 5.1%. In the clinical example, MED3pa identified a patient profile with high misclassification risk, demonstrating its potential for safer deployment. DISCUSSION: By identifying low-confidence predictions, our framework improves model reliability in clinical settings. It can be integrated into decision support systems to help clinicians make more informed decisions. Confidence thresholds help balance model performance with the proportion of patients for whom predictions are considered reliable. CONCLUSION: Better leveraging confidence in model predictions could improve reliability and trustworthiness, supporting safer and more effective use in health care.

Autonomous navigation for wheeled mobile robots generally requires to plan a trajectory and to follow it while performing localization in the environment. In order to execute long-range motion, dead-reckoning localization is not precise enough and the robot must use landmark-based localization. Landmark-based localization can produce discontinuities in the robot position estimation, as it is well known with GPS systems, and perturb the trajectory following. If the robot is able to converge smoothly towards the trajectory, then these perturbations are easily managed. Otherwise, if the robot is more complex (for instance a multi-body mobile robot subject to several nonholonomic constraints) and navigates in a cluttered environment, we show that these perturbations can lead to collisions. The contribution of this paper is to state theoretically the problem and to propose a practical solution to trajectory following for multi-body wheeled mobile robots using landmark-based localization. This solution has been tested on a real robot towing a trailer.

This paper presents a framework for precise parking for nonholonomic mobile robots: the docking task. It consists in following a planned trajectory and reaching a docking configuration, defined relatively to the environment. The trajectory is deformed in order to reach the docking configuration, to avoid obstacles and to keep the nonholonomic constraints satisfied. A generic framework to compute the docking configuration is presented. Then we give the principle of a nonholonomic path deformation method that was used to deform the planned trajectory towards the docking configuration. This framework has been tested on a real robot with a trailer in a realistic scenario

This paper deals with the optimization of Robot-Obstacle interaction computations, in the context of non-holonomic trajectory deformation for mobile robots. We first recall the principle of the trajectory deformation and the role of the potential field gradient in the configuration space. The contribution of the paper is twofold. First we show that the potential field gradient can be computed without any closed-form expression of the potential function if this latter depends only on the distance between the robot and the obstacles. Then an algorithm to filter obstacles that have no influence in Robot-Obstacle interactions is presented. This algorithm takes advantage of the spatial coherence of the planned trajectory, and has been evaluated by experiments on mobile robot Hilare2 towing a trailer.

In this paper we address the problem of obstacles avoidance for car-like robots. We present a generic nonholonomic path deformation method that has been applied on two robots. The principle is to perturb the inputs of the system in order to move away from obstacles and to keep the nonholonomic constraints satisfied. We present an extension of the method to car-like robots. We have integrated the method on two robots (Dala and CyCab) and carried out experiments that show the portability and genericity of the approach.

This paper presents a novel and generic approach of path optimization for nonholonomic systems. The approach is applied to the problem of reactive navigation for nonholonomic mobile robots in highly cluttered environments. This is a collision-free initial path being given for a robot, and obstacles detected while following this path can make it in collision. The current path is iteratively deformed in order to get away from obstacles and satisfy the nonholonomic constraints. The core idea of the approach is to perturb the input functions of the system along the current path in order to modify this path, making an optimization criterion decrease.