Abstract

In this work, we present an algorithm for robot replacement to increase the operational time of a multi-robot payload transport system. Our system comprises a group of non-holonomic wheeled mobile robots traversing on a known trajectory. We design a multi-robot system with loosely coupled robots that ensures the system lasts much longer than the battery life of an individual robot. A system level optimization is presented, to decide on the operational state (charging or discharging) of each robot in the system. The charging state implies that the robot is not in a formation and is kept on charge whereas the discharging state implies that the robot is a part of the formation. Robot battery recharge hubs are present along the trajectory. Robots in the formation can be replaced at these hub locations with charged robots using a replacement mechanism. We showcase the efficacy of the proposed scheduling framework through simulations and experiments with real robots.

Abstract

Literary novels are a mine of information stored in a narrative. Information Extraction (IE) on these novels can be directed by a specific aim to derive knowledge of interest. In this paper, we explore the problem of extracting evidence summaries for all characters in a detective novel. But in this domain, there is no standard annotated text which provides the ground truth for evidence and there exists no fixed metric of evaluation. For extracting summaries, we explore unsupervised methods of learning. We first assume that all characters are equally likely to be culprits and build evidence against each character. Given this evidence, we assess the likelihood of each character being the culprit, eventually arriving at the true culprit(s). Our contribution is the formulation of evidence summaries for characters in determining the culprit. Our experiments are based on a corpus of detective novels spanning the last hundred years. We evaluate these summaries for comprehension with the help of human readers and broadly categorize novels into three groups.

Abstract

In IoT applications, sensors produce raw data, and smart solutions consume the processed data. IoT driven smart solutions typically use data that is aggregated, filtered and processed to drive the smart solution. Such aggregated data has different semantics in comparison to raw sensor data, and it needs to be customized simultaneously for different application requirements. There is a need for proper data logistics for the movement of data from the sensors producing raw data to smart solutions consuming solution relevant data. In this paper, we present a data logistics system that includes (i) architecture to support the movement of data from sensors to smart applications, and (ii) event-driven solution to trigger workflows of smart solution. Workflows in our solution supports the data fulfillment requirements for the enactment of smart solutions

Abstract

High dimensional data is often challenging to cluster due to the curse of dimensionality leading to challenges in identifying clusters. The key challenge in high dimensional clustering is to develop a solution that identifies clusters which are as complete as they can be, while not merging well-separated clusters. We propose core points which represent local compact regions. The strongly connected component from the k-nearest neighbor graph of core points provides for a group of points that are strongly mutually connected. These mutually connected regions represent the core structure of the clusters. Our empirical analysis and experimental results present the rationale behind our solution and validate the goodness of the clusters against the state of the art high dimensional clustering algorithms.The novelty of our solution is to use the concept of reverse nearest neighbors to generate natural clusters in high dimensions.

Abstract

In this work, we consider the problem of decentralized multi-robot target tracking and obstacle avoidance in dynamic environments. Each robot executes a local motion planning algorithm which is based on model predictive control (MPC). The planner is designed as a quadratic program, subject to constraints on robot dynamics and obstacle avoidance. Repulsive potential field functions are employed to avoid obstacles. The novelty of our approach lies in embedding these non-linear potential field functions as constraints within a convex optimization framework. Our method convexifies nonconvex constraints and dependencies, by replacing them as pre-computed external input forces in robot dynamics. The proposed algorithm additionally incorporates different methods to avoid field local minima problems associated with using potential field functions in planning. The motion planner does not enforce predefined trajectories or any formation geometry on the robots and is a comprehensive solution for cooperative obstacle avoidance in the context of multi-robot target tracking. We perform simulation studies for different scenarios to showcase the convergence and efficacy of the proposed algorithm.

Abstract

Always-on agents are those agents which are like daemon programs that are always on and can do tasks as and when they arrive. These agents are idle when there is no task assigned to them. Further, those agents that work on a task, also wait for some event or task completion, and hence, are also idle for a short duration between execution of the tasks. The question arises what should the agent be doing when it is idle. In this paper, we conduct an empirical analysis to show improved decision-making capabilities by agents when they exploit their idle-time to analyze their past tasks in terms of decisions taken and their impact. The improvement of performance can be measured by considering an increase in success rate, avoiding strategies that may not work, quicker decision making, etc. Our always-on agent, while executing a task, stores some of the pertinent details of the task done in a database, such as decisions taken, paths of execution of the task, goodness of them, etc. The agent uses different strategies to use this stored knowledge. We present and evaluate three strategies that always-on agents can use (i) Frequent Decision Strategy (FDS) - the agent stores the prior executions and their frequency of success and failure, repeats the most frequent successful decision taken during prior executions of the task, (ii) Analyzed Decision Strategy (ADS) - the agent analyzes prior executions that were successful or not, stores in database the goodness of various alternatives and chooses the best alternative and (iii) Online analysis decision strategy (OADS) - the alwayson agent while executing its task, during its idle time analyzes the possible future situations and prepares the list of best possible decisions that can be taken in future. Note that the FDS and ADS are used when the agent is not doing any task and is off-line, whereas, OADS generates new mock tasks to consider possible alternative task execution situations to expand its decision-making scenarios while having a task at hand. We conduct our empirical study on always-on agents playing connect-4 games to check the viability and show improved decision making.

Abstract

In the game of hide and seek, one or more agents (hiders) can hide behind objects to prevent other agents (seekers) from seeing them through their line of sight (visibility). These hider agents can be eventually found by seekers who explore the environment. The hide and seek model serves as an abstraction of many real world scenarios such as police agents trying to capture intruders hiding in a warehouse, security/patrolling agents on the lookout for any suspicious activity, spying drones trying to prevent themselves from being caught. Medusa allows modeling such scenarios by supporting the simulation, visualization and analysis of hider and seeker agent strategies under different environments.

Abstract

Mining relationships between treatment(s) and medical problem(s) is vital in the biomedical domain. This helps in various applications, such as decision support system, safety surveillance, and new treatment discovery. We propose a deep learning approach that utilizes both word level and sentence-level representations to extract the relationships between treatment and problem. While deep learning techniques demand a large amount of data for training, we make use of a rule-based system particularly for relationship classes with fewer samples. Our final relations are derived by jointly combining the results from deep learning and rule-based models. Our system achieved a promising performance on the relationship classes of I2b2 2010 relation extraction task.

Abstract

Mining relationships between treatment(s) and medical problem(s) is vital in the biomedical domain. This helps in various applications, such as decision support system, safety surveillance, and new treatment discovery. We propose a deep learning approach that utilizes both word level and sentence-level representations to extract the relationships between treatment and problem. While deep learning techniques demand a large amount of data for training, we make use of a rule-based system particularly for relationship classes with fewer samples. Our final relations are derived by jointly combining the results from deep learning and rule-based models. Our system achieved a promising performance on the relationship classes of I2b2 2010 relation extraction task.

Abstract

We estimate that a large number of news articles contain references to future. The reference is detected through the notion of predictive statements (phrases). Distinguishing such predictive statements from factual statements in news articles is important for most applications such as fact checking, opinion mining, future trend analysis, etc. In this paper, we approach the problem of automatically extracting futurerelated information by solving two sub-problems. The first sub-problem is labeling a sentence as predictive or factual. In addition to extracting the predictions, we address the tasks of clausal scope resolution and dis-embedding linguistic peripheral clauses with respect to the predictive clause in a sentence. To solve these problems, we extract all the clauses of a given sentence and classify each of the clauses as predictive or factual. We then use a machine learning based approach to disambiguate the clause labels by using the clausal dependency relations and label the sentence.