We present and examine a technique for estimating the ego-motion of a mobile robot using memory-based learning and a monocular camera. Unlike other approaches that rely heavily on camera calibration and geometry to compute trajectory, our method learns a mapping from sparse optical flow to platform velocity and turn rate. We also demonstrate an efficient method of computing high-quality sparse optical flow, and techniques for using this sparse optical flow as input to a supervised learning method. We employ a voting scheme of many learners that use subsets of the sparse optical flow to cope with variable dimensionality and reduce the dimensionality of each learner. Finally, we perform experiments in which we examine the learned mapping for visual odometry, investigate the effects of varying the reduced dimensionality of the sparse optical flow state, and quantify the accuracy of two variations of our learner scheme. Our results indicate that our learning scheme estimates monocular visual odometry mainly from points on the ground plane, and reflect to a degree the minimum dimensionality imposed by the problem. In addition, we show that while this memory-based learning method cannot yet estimate ego-motion as accurately as recent geometric methods, it is possible to learn, with no explicit model of camera calibration or scene structure, complicated mappings that take advantage of properties of the camera and the environment.
A new platform for mobile manipulation consisting of a Segway RMP base and a KUKA KR-5 sixx manipulator was developed at the Georgia Institute of Technology in the context of a class. Students formed three teams whose goal was to design a system capable of autonomously serving coffee. Each team took a different approach to the problem in terms of system architecture, visual recognition, and grasping procedure. The approaches used by the students and their merits with respect to this task are presented.
The airborne network (AN) will form an essential part of the global information grid in the future, thus providing information and decision superiority to US armed forces. AN is an enabling technology for network centric warfare. AN is different from the terrestrial mobile ad-hoc networks (MANETs) and the wire-line internet, both in terms of network capability and underlying assumptions. The backbone nodes of the AN are envisioned to fly in pre-planned orbits whose knowledge can be exploited for efficient routing. In this paper we propose a dynamic adaptive routing protocol that uses known trajectories of the AN nodes to enhance performance. Our routing protocol has two components: (1) a Mobility Aware Routing Protocol (MARP), that routes traffic based on the knowledge of network topology with respect to time and makes preemptive decisions to minimize packet losses due to link failure and discover better routes, and (2) a Mobility Dissemination protocol (MDP) that informs all network nodes of any deviation from the preplanned behavior. We analyze MARP/MDP protocol suite using the QualNet network simulator for representative AN deployment scenarios and compare performance with proactive and reactive MANET routing protocols. We use packet delivery ratio, end-to-end latency and control overhead as performance metrics. We also analyze the performance of the MARP/MDP routing protocol for varying degrees of prediction accuracy. This work is part of an ongoing Phase II Small Business Innovation Research program administered by the Air Force Research Laboratory/Information Directorate in Rome, New York.
Active topology management in the future airborne networks (AN) can provide improved overall network throughput, efficiency, and scalability and is critical due to the high degree of platform dynamics involved. The RF links that form an airborne network must be established and reconfigured rapidly in response to aircraft joining and leaving the network, aircraft changing flight paths, and to changes in mission information flows, among other things. Additional technical challenges stem from the fact that the airborne nodes will use multiple directional and omni-directional antennas with differing antenna patterns. In this paper we present a Mobility Aware Topology Control (MAToC) solution for the Airborne Network. MAToC is comprised of deliberative and reactive topology planning components. MAToC utilizes a distributed protocol for airborne nodes for ad-hoc exchange of respective flight plan. Deliberative mode planning uses the collected flight plan information to assign optimal power, channel and boresight direction to the airborne antennas. Deliberative MAToC uses graph coloring algorithms for channel and timeslot assignment and uses geometric optimization methodology to assign antenna powers to maximize Signal to Interference and Noise Ratio (SINR). In the reactive mode, MAToC is responsible for link monitoring and link repair for fault-tolerance.
We have developed a framework, Cognitive Object Recognition System (CORS), inspired by current neurocomputational models and psychophysical research in which multiple recognition algorithms (shape based geometric primitives, 'geons,' and non-geometric feature-based algorithms) are integrated to provide a comprehensive solution to object recognition and landmarking. Objects are defined as a combination of geons, corresponding to their simple parts, and the relations among the parts. However, those objects that are not easily decomposable into geons, such as bushes and trees, are recognized by CORS using "feature-based" algorithms. The unique interaction between these algorithms is a novel approach that combines the effectiveness of both algorithms and takes us closer to a generalized approach to object recognition. CORS allows recognition of objects through a larger range of poses using geometric primitives and performs well under heavy occlusion - about 35% of object surface is sufficient. Furthermore, geon composition of an object allows image understanding and reasoning even with novel objects. With reliable landmarking capability, the system improves vision-based robot navigation in GPS-denied environments. Feasibility of the CORS system was demonstrated with real stereo images captured from a Pioneer robot. The system can currently identify doors, door handles, staircases, trashcans and other relevant landmarks in the indoor environment.
We consider the problem of routing in a mobile ad-hoc network (MANET) for which the planned mobilities of the nodes are partially known a priori and the nodes travel in groups. This situation arises commonly in military and emergency response scenarios. Optimal routes are computed using the most reliable path principle in which the negative logarithm of a node pair's adjacency probability is used as a link weight metric. This probability is estimated using the mobility plan as well as dynamic information captured by table exchanges, including a measure of the social tie strength between nodes. The latter information is useful when nodes deviate from their plans or when the plans are inaccurate. We compare the proposed routing algorithm with the commonly-used optimized link state routing (OLSR) protocol in ns-3 simulations. As the OLSR protocol does not exploit the mobility plans, it relies on link state determination which suffers with increasing mobility. Our simulations show considerably better throughput performance with the proposed approach as compared with OLSR at the expense of increased overhead. However, in the high-throughput regime, the proposed approach outperforms OLSR in terms of both throughput and overhead.
We propose a Markov Decision Process (MDP) approach to optimizing decision making of an Unmanned Aerial Vehicle (UAV) in the presence of communication latency for collision avoidance scenarios. The UAV can be controlled by a remote human pilot or an onboard collision avoidance system (CAS). Due to communication latency, the pilot’s commands may be delayed for several seconds. If the UAV encounters a dynamic intruder, it has to make a decision on whether it should evade immediately by following the safe path generated from the CAS, or wait for the arrival of the pilot’s commands. The proposed MDP will make predictions on the state change in such a scenario. By solving the MDP with the value iteration method, sequential decisions on whether to wait can be made and an optimal waiting strategy can be obtained, which can be used to improve the human pilot’s user experience while ensuring safety.
Mobile manipulation in many respects represents the next generation of robot applications. An important part
of design of such systems is the integration of techniques for navigation, recognition, control, and planning to
achieve a robust solution. To study this problem three different approaches to mobile manipulation have been
designed and implemented. A prototypical application that requires navigation and manipulation has been
chosen as a target for the systems. In this paper we present the basic design of the three systems and draw some
general lessons on design and implementation.
'/%3e%3c/defs%3e%3cpath fill='%23012169' d='M0 0h10080v5040H0z'/%3e%3cpath stroke='%23fff' stroke-width='.6' d='m0 0 6 3m0-3L0 3' clip-path='url(%23b)' transform='scale(840)'/%3e%3cpath stroke='%23e4002b' stroke-width='.4' d='m0 0 6 3m0-3L0 3' clip-path='url(%23c)' transform='scale(840)'/%3e%3cpath stroke='%23fff' stroke-width='840' d='M2520 0v2520M0 1260h5040'/%3e%3cpath stroke='%23e4002b' stroke-width='504' d='M2520 0v2520M0 1260h5040'/%3e%3cg fill='%23fff'%3e%3cuse xlink:href='%23d' x='2520' y='3780'/%3e%3cuse xlink:href='%23a' x='7560' y='4200'/%3e%3cuse xlink:href='%23a' x='6300' y='2205'/%3e%3cuse xlink:href='%23a' x='7560' y='840'/%3e%3cuse xlink:href='%23a' x='8680' y='1869'/%3e%3cuse xlink:href='%23e' x='8064' y='2730'/%3e%3c/g%3e%3c/svg%3e)
'%3e%3ccircle r='20' fill='%23008'/%3e%3ccircle r='17.5' fill='%23fff'/%3e%3ccircle r='3.5' fill='%23008'/%3e%3cg id='d'%3e%3cg id='c'%3e%3cg id='b'%3e%3cg id='a' fill='%23008'%3e%3ccircle r='.875' transform='rotate(7.5 -8.75 133.5)'/%3e%3cpath d='M0 17.5.6 7 0 2l-.6 5L0 17.5z'/%3e%3c/g%3e%3cuse xlink:href='%23a' transform='rotate(15)'/%3e%3c/g%3e%3cuse xlink:href='%23b' transform='rotate(30)'/%3e%3c/g%3e%3cuse xlink:href='%23c' transform='rotate(60)'/%3e%3c/g%3e%3cuse xlink:href='%23d' transform='rotate(120)'/%3e%3cuse xlink:href='%23d' transform='rotate(-120)'/%3e%3c/g%3e%3c/svg%3e)