Planning for a wide range of real-world robotic tasks necessitates to know and write all constraints. However, instances exist where these constraints are either unknown or challenging to specify accurately. A possible solution is to infer the unknown constraints from expert demonstration. This paper presents a novel Positive-Unlabeled Constraint Learning (PUCL) algorithm to infer a continuous arbitrary constraint function from demonstration, without requiring prior knowledge of the true constraint parameterization or environmental model as existing works. Within our framework, we treat all data in demonstrations as positive (feasible) data, and learn a control policy to generate potentially infeasible trajectories, which serve as unlabeled data. In each iteration, we first update the policy and then a two-step positive-unlabeled learning procedure is applied, where it first identifies reliable infeasible data using a distance metric, and secondly learns a binary feasibility classifier (i.e., constraint function) from the feasible demonstrations and reliable infeasible data. The proposed framework is flexible to learn complex-shaped constraint boundary and will not mistakenly classify demonstrations as infeasible as previous methods. The effectiveness of the proposed method is verified in three robotic tasks, using a networked policy or a dynamical system policy. It successfully infers and transfers the continuous nonlinear constraints and outperforms other baseline methods in terms of constraint accuracy and policy safety.

Planning for a wide range of real-world tasks necessitates to know and write all constraints. However, instances exist where these constraints are either unknown or challenging to specify accurately. A possible solution is to infer the unknown constraints from expert demonstration. The majority of prior works limit themselves to learning simple linear constraints, or require strong knowledge of the true constraint parameterization or environmental model. To mitigate these problems, this paper presents a positive-unlabeled (PU) learning approach to infer a continuous, arbitrary and possibly nonlinear, constraint from demonstration. From a PU learning view, We treat all data in demonstrations as positive (feasible) data, and learn a (sub)-optimal policy to generate high-reward-winning but potentially infeasible trajectories, which serve as unlabeled data containing both feasible and infeasible states. Under an assumption on data distribution, a feasible-infeasible classifier (i.e., constraint model) is learned from the two datasets through a postprocessing PU learning technique. The entire method employs an iterative framework alternating between updating the policy, which generates and selects higher-reward policies, and updating the constraint model. Additionally, a memory buffer is introduced to record and reuse samples from previous iterations to prevent forgetting. The effectiveness of the proposed method is validated in two Mujoco environments, successfully inferring continuous nonlinear constraints and outperforming a baseline method in terms of constraint accuracy and policy safety.

Combinatorial problems are ubiquitous in the world around us. Actually, they are found in all fields of human activity. As illustrations, it may be a question of scheduling the operations to be carried out within an industrial process (production line of a vehicle, an airplane or a satellite), of extracting the recurring patterns in a transaction database (data mining), of organizing the roster of a service (in a hospital, university or industrial environment), of generating molecular structures with good properties (in chemistry or bioinformatics), etc. Solving optimization problems remains a difficult task, especially when the size of the instances of the problems to be solved is large and/or when optimality is desired. In reality, the difficulty is twofold: being able to appropriately write models for encountered problems and being able to effectively solve the different instances of these problems. The main paradigms for optimization, namely mathematical programming, metaheuristics and Constraint Programming (CP), including the Boolean SAT framework, offer varied and interesting tools (languages, libraries, software), and are in a way, quite complementary; each paradigm having its own success stories.

In this paper, we present an efficient algorithm for verifying path-consistency on a binary constraint network. The complexities of our algorithm beat the previous conjectures on the lower bounds for verifying path-consistency. We therefore defeat the proofs for several published results that incorrectly rely on these conjectures. Our algorithm is motivated by the idea of reformulating path-consistency verification as fast matrix multiplication. Further, for a computational model that counts arithmetic operations (rather than bit operations), a clever use of the properties of prime numbers allows us to design an even faster variant of the algorithm. Based on our algorithm, we hope to inspire a new class of techniques for verifying and even establishing varying levels of local-consistency on given constraint networks.

The size of these networks is quadratic in the number of variables, which has severely limited the real-world application of QSR. In this paper, we propose another representation of spatial scenarios, in which each variable is associated with one or more rectangles. Instead of requiring these rectangles to define a solution of the corresponding constraint network, we construct sequences of rectangles that define partial solutions to progressively weaker constraint networks. We present experimental results that illustrate the effectiveness of this strategy.

In most optimization problems, users have a clear understanding of the function to optimize (e.g., minimize the makespan for scheduling problems). However, the constraints may be difficult to state and their modelling often requires expertise in Constraint Programming. Active constraint acquisition has been successfully used to support non-experienced users in learning constraint networks through the generation of a sequence of queries. In this paper, we propose Learn&Optimize, a method to solve optimization problems with known objective function and unknown constraint network. It uses an active constraint acquisition algorithm which learns the unknown constraints and computes boundaries for the optimal solution during the learning process. As a result, our method allows users to solve optimization problems without learning the overall constraint network.

Constraint acquisition systems such as QuAcq and MultiAcq can assist non-expert users to model their problems as constraint networks by classifying (partial) examples as positive or negative. For each negative example, the former focuses on one constraint of the target network, while the latter can learn a maximum number of constraints. Two bottlenecks of the acquisition process where both these algorithms encounter problems are the large number of queries required to reach convergence, and the high cpu times needed to generate queries, especially near convergence. In this paper we propose algorithmic and heuristic methods to deal with both these issues. We first describe an algorithm, called MQuAcq, that blends the main idea of MultiAcq into QuAcq resulting in a method that learns as many constraints as MultiAcq does after a negative example, but with a lower complexity. A detailed theoretical analysis of the proposed algorithm is also presented. %We also present a technique that boosts the performance of constraint acquisition by reducing the number of queries significantly. Then we turn our attention to query generation which is a significant but rather overlooked part of the acquisition process. We describe %in detail how query generation in a typical constraint acquisition system operates, and we propose heuristics for improving its efficiency. Experiments from various domains demonstrate that our resulting algorithm that integrates all the new techniques does not only generate considerably fewer queries than QuAcq and MultiAcq, but it is also by far faster than both of them, in average query generation time as well as in total run time, and also largely alleviates the premature convergence problem.

Generating a novel and optimized molecule with desired chemical properties is an essential part of the drug discovery process. Failure to meet one of the required properties can frequently lead to failure in a clinical test which is costly. In addition, optimizing these multiple properties is a challenging task because the optimization of one property is prone to changing other properties. In this paper, we pose this multi-property optimization problem as a sequence translation process and propose a new optimized molecule generator model based on the Transformer with two constraint networks: property prediction and similarity prediction. We further improve the model by incorporating score predictions from these constraint networks in a modified beam search algorithm. The experiments demonstrate that our proposed model outperforms state-of-the-art models by a significant margin for optimizing multiple properties simultaneously.

It is not usual to be asked to write something on a subject on which you have worked fifteen years before and on which you have remained far from subsequent developments. Yet this is the challenge we humbly accepted to undertake in this paper with the limitation that our expertise in this research area have diminished regarding the developments made by many others during the ten previous years. We begin with the easy part consisting in recalling the context of the paper [BB03]. In the sequel the word constraint will be used in two meanings: a type of constraints, such as propositional clause, Boolean cardinality constraint, ALLDIFF..., and a constraint instance, i.e., the formal representation of a relation on several variables each having a domain. We will use the word Constraint in the first case and constraint in the second case.

Learning constraint networks is known to require a number of membership queries exponential in the number of variables. In this paper, we learn constraint networks by asking the user partial queries. That is, we ask the user to classify assignments to subsets of the variables as positive or negative. We provide an algorithm, called QUACQ, that, given a negative example, focuses onto a constraint of the target network in a number of queries logarithmic in the size of the example. The whole constraint network can then be learned with a polynomial number of partial queries. We give information theoretic lower bounds for learning some simple classes of constraint networks and show that our generic algorithm is optimal in some cases.