Abhishek Anand

Contextually guided semantic labeling and search for three-dimensional point clouds

RGB-D cameras, which give an RGB image together with depths, are becoming increasingly popular for robotic perception. In this paper, we address the task of detecting commonly found objects in the three-dimensional (3D) point cloud of indoor scenes obtained from such cameras. Our method uses a graphical model that captures various features and contextual relations, including the local visual appearance and shape cues, object co-occurrence relationships and geometric relationships. With a large number of object classes and relations, the model’s parsimony becomes important and we address that by using multiple types of edge potentials. We train the model using a maximum-margin learning approach. In our experiments concerning a total of 52 3D scenes of homes and offices (composed from about 550 views), we get a performance of 84.06% and 73.38% in labeling office and home scenes respectively for 17 object classes each. We also present a method for a robot to search for an object using the learned model and the contextual information available from the current labelings of the scene. We applied this algorithm successfully on a mobile robot for the task of finding 12 object classes in 10 different offices and achieved a precision of 97.56% with 78.43% recall.1

ROSCoq: Robots Powered by Constructive Reals

We present ROSCoq, a framework for developing certified Coq programs for robots. ROSCoq subsystems communicate using messages, as they do in the Robot Operating System (ROS). We extend the logic of events to enable holistic reasoning about the cyber-physical behavior of robotic systems. The behavior of the physical world (e.g. Newton’s laws) and associated devices (e.g. sensors, actuators) are specified axiomatically. For reasoning about physics we use and extend CoRN’s theory of constructive real analysis. Instead of floating points, our Coq programs use CoRN’s exact, yet fast computations on reals, thus enabling accurate reasoning about such computations.

Towards a Formally Verified Proof Assistant

This paper presents a formalization of Nuprl’s metatheory in Coq. It includes a nominal-style definition of the Nuprl language, its reduction rules, a coinductive computational equivalence, and a Curry-style type system where a type is defined as a Partial Equivalence Relation (PER) a la Allen. This type system includes Martin-Lof dependent types, a hierarchy of universes, inductive types and partial types. We then prove that the typehood rules of Nuprl are valid w.r.t. this PER semantics and hence reduce Nuprl’s consistency to Coq’s consistency.

Formal program optimization in nuprl using computational equivalence and partial types

This paper extends the proof methods used by the Nuprl proof assistant to reason about the computational behavior of its untyped programs. We have implemented new methods to prove non-trivial bisimulations between programs and have successfully applied these methods to formally optimize distributed programs such as our synthesized and verified version of Paxos, a widely used protocol to achieve software based replication. We prove new results about the basic computational equality relation on terms, and we extend the theory of partial types as the basis for stating internal results about the computation system that were previously treated only in the meta theory of Nuprl. All the lemmas presented in this paper have been formally proved in Nuprl.

Towards Certified Meta-Programming with Typed Template-Coq

Template-Coq is a plugin for Coq, originally implemented by Malecha, which provides a reifier for Coq terms and global declarations , as represented in the Coq kernel, as well as a denotation command. Initially, it was developed for the purpose of writing functions on Coqs AST in Gallina. Recently, it was used in the CertiCoq certified compiler project, as its front-end language, to derive parametricity properties, and to extract Coq terms to a CBV λ-calculus. However, the syntax lacked semantics, be it typing semantics or operational semantics, which should reflect, as formal specifications in Coq, the semantics of Coqs type theory itself. The tool was also rather bare bones, providing only rudimentary quoting and unquoting commands. We generalize it to handle the entire Calculus of Inductive Constructions (CIC), as implemented by Coq, including the kernels declaration structures for definitions and inductives, and implement a monad for general manipulation of Coqs logical environment. We demonstrate how this setup allows Coq users to define many kinds of general purpose plugins, whose correctness can be readily proved in the system itself, and that can be run efficiently after extraction. We give a few examples of implemented plugins, including a parametricity translation. We also advocate the use of Template-Coq as a foundation for higher-level tools.