Deep learning has rapidly transformed the state of the art algorithms used to address a variety of problems in computer vision and robotics. These breakthroughs have relied upon massive amounts of human annotated training data. This time consuming process has begun impeding the progress of these deep learning efforts. This paper describes a method to incorporate photo-realistic computer images from a simulation engine to rapidly generate annotated data that can be used for the training of machine learning algorithms. We demonstrate that a state of the art architecture, which is trained only using these synthetic annotations, performs better than the identical architecture trained on human annotated real-world data, when tested on the KITTI data set for vehicle detection. By training machine learning algorithms on a rich virtual world, real objects in real scenes can be learned and classified using synthetic data. This approach offers the possibility of accelerating deep learning's application to sensorbased classification problems like those that appear in selfdriving cars. The source code and data to train and validate the networks described in this paper are made available for researchers.

Abstract-In this paper, we present an approach for designing feedback controllers for polynomial systems that maximize the size of the time-limited backwards reachable set (BRS). We rely on the notion of occupation measures to pose the synthesis problem as an infinite dimensional linear program (LP) and provide finite dimensional approximations of this LP in terms of semidefinite programs (SDPs). The solution to each SDP yields a polynomial control policy and an outer approximation of the largest achievable BRS. In contrast to traditional Lyapunov based approaches, which are non-convex and require feasible initialization, our approach is convex and does not require any form of initialization. The resulting time-varying controllers and approximated backwards reachable sets are well-suited for use in a trajectory library or feedback motion planning algorithm. We demonstrate the efficacy and scalability of our approach on four nonlinear systems.

No abstract

OBJECTIVE-Increases in brain cyclooxygenase-2 (COX2) are associated with the central inflammatory response and with delayed neuronal death, events that cause secondary insults after traumatic brain injury. A growing literature supports the benefit of COX2-specific inhibitors in treating brain injuries. [5,5-dimethyl-3(3-fluorophenyl)-4(4-methylsulfonyl)phenyl-2( 5 H)-furanone] is a third-generation, highly specific COX2 enzyme inhibitor. DFU treatments (1 or 10 mg/kg intraperitoneally, twice daily for 3 d) were initiated either before or after traumatic brain injury in a lateral cortical contusion rat model. METHODS-DFU NIH-PA Author ManuscriptNIH-PA Author Manuscript NIH-PA Author Manuscript RESULTS-DFU treatments initiated 10 minutes before injury or up to 6 hours after injury enhanced functional recovery at 3 days compared with vehicle-treated controls. Significant improvements in neurological reflexes and memory were observed. DFU initiated 10 minutes before injury improved histopathology and altered eicosanoid profiles in the brain. DFU 1 mg/kg reduced the rise in prostaglandin E 2 in the brain at 24 hours after injury. DFU 10 mg/kg attenuated injuryinduced COX2 immunoreactivity in the cortex (24 and 72 h) and hippocampus (6 and 72 h). This treatment also decreased the total number of activated caspase-3-immunoreactive cells in the injured cortex and hippocampus, significantly reducing the number of activated caspase-3-immunoreactive neurons at 72 hours after injury. DFU 1 mg/kg amplified potentially anti-inflammatory epoxyeicosatrienoic acid levels by more than fourfold in the injured brain. DFU 10 mg/kg protected the levels of 2-arachidonoyl glycerol, a neuro-protective endocannabinoid, in the injured brain.CONCLUSION-These improvements, particularly when treatment began up to 6 hours after injury, suggest exciting neuroprotective potential for COX2 inhibitors in the treatment of traumatic brain injury and support the consideration of Phase I/II clinical trials.Keywords 2-Arachidonoyl glycerol; Caspase-3; COX2 inhibitor; Eicosanoids; Endocannabinoid; Neuroinflammation; Rat behavior Traumatic brain injury (TBI) initiates a central inflammatory response that results in the increased production of prostaglandins and reactive oxygen species. These products may cause secondary insults to the brain (55,90,94). Cyclooxygenases catalyze the first step in the formation of prostaglandins from arachidonic acid, producing reactive oxygen species in the process. Cyclooxygenase-1 (COX1), present normally in most tissues, is thought to maintain essential physiological functions, such as gastric mucosal integrity, renal function, and platelet homeostasis (87,101). COX2, the inducible isoform, is expressed in the normal brain (30,87) but is rare or absent in most organs under normal conditions. As in the periphery (32,79), COX2 in the brain can be regulated by inflammatory cytokines (87), and prostaglandin production increases in response to pathological conditions of the central nervous system (3,70,72,89,107). ...

Mechanistically linking movement behaviors and ecology is key to understanding the adaptive evolution of locomotion. Predator evasion, a behavior that enhances fitness, may depend upon short bursts or complex patterns of locomotion. However, such movements are poorly characterized by existing biomechanical metrics. We present methods based on the entropy measure of randomness from Information Theory to quantitatively characterize the unpredictability of non-steady-state locomotion. We then apply the method by examining sympatric rodent species whose escape trajectories differ in dimensionality. Unlike the speed-regulated gait use of cursorial animals to enhance locomotor economy, bipedal jerboa (family Dipodidae) gait transitions likely enhance maneuverability. In field-based observations, jerboa trajectories are significantly less predictable than those of quadrupedal rodents, likely increasing predator evasion ability. Consistent with this hypothesis, jerboas exhibit lower anxiety in open fields than quadrupedal rodents, a behavior that varies inversely with predator evasion ability. Our unpredictability metric expands the scope of quantitative biomechanical studies to include non-steady-state locomotion in a variety of evolutionary and ecologically significant contexts.

No abstract

No abstract

Controlling soft robots with precision is a challenge due in large part to the difficulty of constructing models that are amenable to model-based control design techniques. Koopman operator theory offers a way to construct explicit linear dynamical models of soft robots and to control them using established model-based linear control methods. This method is data-driven, yet unlike other data-driven models such as neural networks, it yields an explicit control-oriented linear model rather than just a "black-box" input-output mapping. This work describes this Koopman-based system identification method and its application to model predictive controller design. A model and MPC controller of a pneumatic soft robot arm is constructed via the method, and its performance is evaluated over several trajectory following tasks in the real-world. On all of the tasks, the Koopman-based MPC controller outperforms a benchmark MPC controller based on a linear state-space model of the same system.