An effective hybrid search mode for multi-objective optimization with constraints

In this paper a new search strategy for multi-objective optimization (MOO) with constraints is proposed based on a hybrid search mode (HSM). The search processes for feasible solutions and optimal solutions are executed in a mixed way for the existing methods. With regard to HSM, a hybrid search mode is proposed, which consists of two processes: Feasibility search mode (FSM) and optimal search mode (OSM). The executions of these two search modes are independent relatively and also adjusted according to the population distribution. In the early stage, FSM plays the leading role for exploring the feasible space since most of the individuals are infeasible. With the increase of the feasible individuals, OSM is the primary operation for the search of optimal individuals. The proposed method is simple to implement and need few extra parameter tuning. The handing method of constraints is tested on several multi-objective optimization problems with constraints. The remarkable results demonstrate its effectiveness and good performance.

Representation, learning, and planning algorithms for geometric task and motion planning

We present a framework for learning to guide geometric task-and-motion planning (G-TAMP). G-TAMP is a subclass of task-and-motion planning in which the goal is to move multiple objects to target regions among movable obstacles. A standard graph search algorithm is not directly applicable, because G-TAMP problems involve hybrid search spaces and expensive action feasibility checks. To handle this, we introduce a novel planner that extends basic heuristic search with random sampling and a heuristic function that prioritizes feasibility checking on promising state–action pairs. The main drawback of such pure planners is that they lack the ability to learn from planning experience to improve their efficiency. We propose two learning algorithms to address this. The first is an algorithm for learning a rank function that guides the discrete task-level search, and the second is an algorithm for learning a sampler that guides the continuous motion-level search. We propose design principles for designing data-efficient algorithms for learning from planning experience and representations for effective generalization. We evaluate our framework in challenging G-TAMP problems, and show that we can improve both planning and data efficiency.

Semantic Linking Maps for Active Visual Object Search (Extended Abstract)

We aim for mobile robots to function in a variety of common human environments, which requires them to efficiently search previously unseen target objects. We can exploit background knowledge about common spatial relations between landmark objects and target objects to narrow down search space. In this paper, we propose an active visual object search strategy method through our introduction of the Semantic Linking Maps (SLiM) model. SLiM simultaneously maintains the belief over a target object's location as well as landmark objects' locations, while accounting for probabilistic inter-object spatial relations. Based on SLiM, we describe a hybrid search strategy that selects the next best view pose for searching for the target object based on the maintained belief. We demonstrate the efficiency of our SLiM-based search strategy through comparative experiments in simulated environments. We further demonstrate the real-world applicability of SLiM-based search in scenarios with a Fetch mobile manipulation robot.

An Ensemble of Hybrid Search-Based Algorithms for Software Effort Prediction

Software organizations rely on the estimation of efforts required for the development of software to negotiate customers and plan the schedule of the project. Proper estimation of efforts reduces the chances of project failures. Historical data of projects have been used to predict the effort required for software development. In recent years, various ensemble of machine learning techniques have been used to predict software effort. In the present work, a novel ensemble technique of hybrid search-based algorithms (EHSBA) is used for software effort estimation. Four HSBAs—fuzzy and random sets-based modeling (FRSBM-R), symbolic fuzzy learning based on genetic programming (GFS-GP-R), symbolic fuzzy learning based on genetic programming grammar operators and simulated annealing (GFS_GSP_R), and least mean squares linear regression (LinearLMS_R)—are used to create an ensemble (EHSBA). The EHSBA is compared with machine learning-based ensemble bagging, vote, and stacking on datasets obtained from PROMISE repository. Obtained results reported that EHSBA outperformed all other techniques.

Robot calibration based on nonlinear formulation for modular reconfigurable robots (MRRs).

Developed in this thesis is a full pose kinematic calibration method for modular reconfigurable robots (MRRs). This method is based on a nonlinear formulation as opposed to the conventional linear method that has a number of critical limitations. By avoiding linearization of the nonlinear robot forward kinematic equations, these nonlinear equations are directly used to identify the robot calibration parameters. A hybrid search method is developed to solve the nonlinear error equations by combining genetic algorithms with Monte Carlo simulations to ensure a global search over the robot workspace with good accuracy. A number of comparisons are made between the proposed method and the conventional linear method, indicating the advantages of the former over the latter by eliminating two critical limitations. The first one is the orthogonality sacrifice that leads to ill-conditioning of the Jacobian used in the linear method. The second one is quadrant sensitivity during the determination of the (Tait) Bryan angles from inverting the rotation matrix.