December 2024
What is this page?
This page lists works of an author who doesn't have a ResearchGate profile or hasn't added the works to their profile yet. It is automatically generated from public (personal) data to further our legitimate goal of comprehensive and accurate scientific recordkeeping. If you are this author and want this page removed, please let us know.
Publications (47)
December 2024
December 2024
·
4 Reads
We propose MMD-OPT: a sample-efficient approach for minimizing the risk of collision under arbitrary prediction distribution of the dynamic obstacles. MMD-OPT is based on embedding distribution in Reproducing Kernel Hilbert Space (RKHS) and the associated Maximum Mean Discrepancy (MMD). We show how these two concepts can be used to define a sample efficient surrogate for collision risk estimate. We perform extensive simulations to validate the effectiveness of MMD-OPT on both synthetic and real-world datasets. Importantly, we show that trajectory optimization with our MMD-based collision risk surrogate leads to safer trajectories at low sample regimes than popular alternatives based on Conditional Value at Risk (CVaR).
December 2024
·
1 Read
IEEE Robotics and Automation Letters
We propose a learned probabilistic neural policy for safe, occlusion-free target tracking. The core novelty of our work stems from the structure of our policy network that combines generative modeling based on Conditional Variational Autoencoder (CVAE) with differentiable optimization layers. The weights of the CVAE network and the parameters of the differentiable optimization can be learned in an end-to-end fashion through demonstration trajectories. We improve the state-of-the-art (SOTA) in the following respects. We show that our learned policy outperforms existing SOTA in terms of occlusion/collision avoidance capabilities and computation time. Second, we present an extensive ablation showing how different components of our learning pipeline contribute to the overall tracking task. We also demonstrate the real-time performance of our approach on resource-constrained hardware such as NVIDIA Jetson TX2. Finally, our learned policy can also be viewed as a reactive planner for navigation in highly cluttered environments.
November 2024
·
2 Reads
November 2024
·
2 Reads
![Fig. 2: Overview of the Proposed Method: (a) The pipeline shows how the policy network uses observations Ot, reward Rt, time, and mass embeddings to predict stiffness K. This stiffness, along with state variables from MuJoCo and reference trajectory, is fed to the QP solver, which outputs joint accelerations¨qaccelerations¨ accelerations¨q ⋆ L and¨qand¨ and¨q ⋆ R . These accelerations are then converted to torques τ L and τ R and applied to the MuJoCo simulator. (b) Illustrates the QP solver implemented using CVXPY [12], which computes impedance and postural errors using the provided stiffness and reference trajectory and solves the optimization problem (Equation 5) with constraints (Equations 6, 7, 8, 9) to determine the joint accelerations. (c) Depicts the trajectory generation process, which uses a quintic polynomial for cartesian positions and SLERP for SO(3) orientations. The trajectory includes a waypoint where the x and y coordinates are the averages of the initial and goal positions, and the z coordinate is set to a random height within the workspace of the arms to introduce variability.](https://www.researchgate.net/publication/385292238/figure/fig1/AS:11431281286533699@1730085999832/Overview-of-the-Proposed-Method-a-The-pipeline-shows-how-the-policy-network-uses_Q320.jpg)
October 2024
·
92 Reads
Dual-arm manipulation is an area of growing interest in the robotics community. Enabling robots to perform tasks that require the coordinated use of two arms, is essential for complex manipulation tasks such as handling large objects, assembling components, and performing human-like interactions. However, achieving effective dual-arm manipulation is challenging due to the need for precise coordination, dynamic adaptability, and the ability to manage interaction forces between the arms and the objects being manipulated. We propose a novel pipeline that combines the advantages of policy learning based on environment feedback and gradient-based optimization to learn controller gains required for the control outputs. This allows the robotic system to dynamically modulate its impedance in response to task demands, ensuring stability and dexterity in dual-arm operations. We evaluate our pipeline on a trajectory-tracking task involving a variety of large, complex objects with different masses and geometries. The performance is then compared to three other established methods for controlling dual-arm robots, demonstrating superior results.
October 2024
·
18 Reads
International Journal of Biological Macromolecules
August 2024
·
10 Reads
August 2024
·
9 Reads
Citations (29)
... Figure 3a shows the dual-swept transfer curves of the device under different V IN sweep widths with fixed V DS , exhibiting a convex-shaped anti-ambipolar transfer characteristics, where the channel's conductance peaks at a specific V IN . [50] Figure 3b,c exhibit the reverse swept I DS -V IN characteristics extracted from Figure 3a in logarithmic scale and contour chart, respectively, to give a better representation to show the conversion process between the n-type FeFET mode and the p-type MT mode along V IN . Figure 3d shows the corresponding I GS with an obvious boundary at V IN of 1 V, which was considered to divide the I DS -V IN curves from FET mode to MT mode. ...
- Citing Article
July 2024
Nano Letters
... Chalcogenide perovskites hold promise as potential solar cell absorbers due to their comparable electronic and optical properties [23][24][25][26]. These properties include direct band gaps and similar structures. ...
- Citing Article
June 2024
Progress in Solid State Chemistry
... This strategy combines MPPI with a Sparse Gaussian Process (SGP)-based local perception model, incorporating online learning to effectively explore the surrounding navigable space. Meanwhile, in [25], the authors introduce Projection Guided Sampling Based Optimization (PRIEST), a new optimizer designed to push the infeasible sampled trajectories of MPPI towards feasible regions, thereby reducing the risk of local minima and infeasible control sequences. ...
- Citing Article
March 2024
IEEE Robotics and Automation Letters
... In other words, one way of reducing risk is to minimize the difference between p h and p δ . Thus, we propose the following risk estimate following our prior works [15], [16]. ...
- Citing Conference Paper
- Full-text available
October 2023
... Our work extends [18] to the non-convex multi-robot trajectory optimization setting. Contribution Over Author's Prior Works: The proposed work extends [24], [25] to the multi-robot setting. Moreover, our SF solver is an improved and batched version of the optimizer presented in [10]. ...
- Citing Conference Paper
October 2023
... For the case of dynamic target, Manoharan et al. [18] proposed a nonlinear model predictive control-based framework to deal with the three-agent TAD game, where the target-defender team acts in a cooperative manner and an extended Kalman filter is adopted to estimate the states of the attacker. Moreover, the games with two-target two-attacker [19] and multi-defender multi-attacker [20] are also investigated in depth. ...
- Citing Conference Paper
June 2023
... Lin et al [34] addressed the problem of platform vibration under wave loading by obtaining nonlinear stochastic wave loads via Morison's equation and providing statistical moments of the stochastic response via Monte Carlo simulation. Prakash et al [35][36][37] studied the thermoelectric effects of graphene structures using TCAD simulations. ...
- Citing Article
- Publisher preview available
December 2021
Journal of Computational Electronics
... In recent years, graphene, a remarkable two-dimensional (2D) material, has garnered significant attention in photodetection due to its high carrier mobility, sensitivity, current carrying capacity, tunable Fermi level, low resistivity, and broadband light absorption [8,9]. Consequently, graphene has emerged as a potentially valuable material for constructing suitable heterojunctions when combined with other semiconductor materials in photodetector devices [4,[10][11][12][13][14][15][16][17]. These material heterojunctions provide a high internal electric field, an advantageous trait for the rapid separation of photogenerated charge carriers without requiring any external bias voltage. ...
- Citing Chapter
September 2021
... The transistor densities on a single chip are increased by the low operational voltage requirements and fast processing rate of short channel MOS. Even with all these benefits, short channel devices have significant drawbacks [29][30][31][32][33][34]. A few of which are briefly discussed below: ...
- Citing Article
- Full-text available
February 2021
... In recent years, substantial efforts have been directed toward investigating the potential of plasmonic metamaterial absorbers for a diverse array of applications, including plasmonic detectors [25,26] and filters [27]. Following the groundbreaking demonstration of a metamaterial absorber functioning at microwave wavelengths-utilizing an electric ring resonator in conjunction with a cut-wire media structure-numerous metal-insulator-metal (MIM) plasmonic absorbers have been developed [28,29]. While most MIMbased plasmonic absorbers typically feature straightforward lattice configurations in the top metal layer, such as onedimensional or square lattice designs, aimed at achieving significant absorption through their ground state localized surface plasmons [5], there has been comparatively less emphasis on higher-order states or their interactions within particle networks. ...
- Citing Article
January 2021
Journal of Electronic Materials
