SuperPADL: Scaling Language-Directed Physics-Based Control with Progressive Supervised Distillation

This chapter explores creating a physics-based character and moving the character using physics forces. Basing a character on a physics object allows it to interact with any other physics objects in the game scene. Artists will be creating a character, picking an appropriate collider shape to simulate its interactions, and giving it a rigidbody—which allows it to have mass, drag, and gives artists physical forces something to interact with. In the code, they have created a character called robot friend who is moving around a square space and gathering spheres. Artists added a simple robot model to their scene hierarchy and named it RobotFriend. Next, they need to add a RigidBody component to their RobotFriend. The Rigidbody component is responsible for simulating the physical forces: gravity, drag, and giving our game object mass. Next, they need to add a collider component. They selected a Capsule Collider component and added it to their RobotFriend game object.

Homogeneous charge compression ignition (HCCI) is a novel combustion strategy for IC engines that exhibits dramatic decreases in fuel consumption and exhaust emissions. Originally conceived in 1979, the HCCI methodology has been revisited several times by industry but has yet to be implemented because the process is difficult to control. To help address these control challenges, the authors here outline the first generalizable, validated, and experimentally implemented physics-based control methodology for residual-affected HCCI engines. Specifically, the paper describes the formulation and validation of a two-input, two-state control-oriented system model of the residual-affected HCCI process occurring in a single engine cylinder. The combustion timing and peak pressure are the model states, while the inducted gas composition and effective compression ratio are the model inputs. The resulting model accurately captures the system dynamics and allows the simultaneous, coordinated control of both in-cylinder pressure and combustion timing. To demonstrate this, an H2 optimal controller is synthesized from a linearized version of the model and used to dictate step changes in both combustion timing and peak pressure within about four to five engine cycles on an experimental test bed. The application of control also results in reductions in the standard deviation for both combustion timing and peak pressure. The approach therefore provides accurate mean tracking, as well as a reduction in cyclic dispersion. Another benefit of the simultaneous coordination of both control inputs is a reduction in the control effort required to elicit the desired response. Instead of using a peak pressure controller that must compensate for the effects of a combustion timing controller, and vice versa, the coordinated approach optimizes the use of both control inputs to regulate both outputs.

SCR on Filter (SCRoF) is an efficient and compact NOX and PM reduction technology already used in series production for light-duty applications. The technology is now finding its way into the medium duty and heavy duty market. One of the key challenges for successful application is the robustness to real world variations. The solution to this challenge can be found by using model-based control algorithms, utilizing state estimation by physics-based catalyst models. This paper focuses on the development, validation and real time implementation of a physics-based control oriented SCRoF model. An overview of the developed model will be presented, together with a brief description of the model parameter identification and validation process using engine test bench measurement data. The model parameters are identified following a streamlined approach, focusing on decoupling the effects of deNOx and soot phenomena. Model validation results will be shown demonstrating good model performance for both steady state and transient conditions. Following the validation of the off-line SCRoF model, the model was implemented on a dSPACE platform to demonstrate its real-time capability. The implementation on the dSPACE system is verified by comparing the online model signal outputs to the offline model. Further on, the model is compiled into object code and will be integrated into the ECM. The embedded model is characterized by a low memory footprint and a fast turnaround time.

Summary form only given. Following the Northeast blackout of 2003, the Wide-Area Measurement Systems (WAMS) technology using Phasor Measurement Units (PMUs) has tremendously matured for the North American grid. However, with the rapidly increasing number of deployed PMUs, the design and deployment of an efficient wide-area communication and computing infrastructure remains one of the greatest challenges to harvest the gigantic volumes of PMU data in real-time. Neither the architecture of such networks nor the impacts of their operational uncertainties such as delays and data losses on the envisioned control actions are well understood in the current state of art. This tutorial talk will address this gap, and propose a highly resilient, fault-tolerant, and reliable distributed network control system for tomorrow's power grids using cutting-edge emerging technologies such as cloud computing and software defined networks. The proposed architecture will be realized via three interactive layers. Layer 1 will consist of physics-based controllers for power oscillation damping. Layer 2 will contain delay control rules for the communication network that work in tandem with the grid controllers. Layer 3 will consist of a supervisory controller realized through embedding and reconfiguration rules in a distributed cloud environment that continuously monitors the system status, and ensures fault-tolerance, resilience, and reliability of the overall closed-loop control system.

Real-Time Control of Tokamak Plasmas: from Control of Physics to Physics-Based Control

Physics-based control of virtual characters traditionally uses high simulation frequencies of 1 to 2 kHz. While lowering the simulation frequency frees computation time, it usually introduces instabilities within the simulation. In this paper, we propose a control strategy that can be used for high and low simulation frequencies, down to 225 Hz. The inherent instabilities were reduced by optimizing control parameters and by introducing a novel control feedback for the stance leg. We also show how lower frequencies hold a more restrictive space of possible control parameters than higher ones. Our controller shows equal robustness as high frequency controllers while requiring in average only 0.8 ms per simulation step.

Due to the nature of their operations, electronic data centers are highly dynamic in their computational load handling, and concomitantly in the heat generated by various servers inside the data center. In the absence of physics-based control algorithms, air-conditioning systems are typically operated through conservatively pre-determined set points, thereby resulting in sub-optimal electricity consumption. Computational fluid dynamics based models that have been developed are too computationally intensive to be used in control algorithms. Here, we develop a validated transient model for a data center that can be used as the basis for model predictive control algorithms. We consider a common raised floor plenum data center in which cold air is delivered to the server racks and hot air is exhausted through the ceiling. We test and demonstrate the performance of the data driven model on this data center. Thermal and flow dynamics in a data center are strongly affected by rack inlet and rack outlet temperatures, in addition to the computational load (and thereby the heat generated by the racks). Using these variables, the response data were collected at the top, center and bottom position of server simulator rack inside a raised floor plenum data center. The server simulator rack consists of four server simulators, each with independent control of heat load & fan speed. The simulator was first allowed to reach steady state (by holding conditions constant for over 3 hours). Subsequently, the heat load was ramped up in steps. To reach steady state, all simulators were held at 2.5 kW constant heat output with 0.15 m3/s fan speed. The computer room air-conditioning (CRAC) unit was modeled as a constant air supply at 17 °C. The response data were collected for various cases with heat load as input to the system. This input-output data was used for data driven modeling to predict the rack outlet air temperatures. The data analytics techniques that were used include partial correlation and total least square approach. Results show good agreement with experiments and have prediction capacity of 15 minutes. These results indicate that data driven modeling can potentially be used to form the basis for model predictive control.

Spacecraft control suffers from inter-axis coupling regardless of control methodology due to the physics that dominate their motion.Feedback control is used to robustly reject disturbances, but is complicated by this coupling.Other sources of disturbances include zero-virtual references associated with cascaded control loop topology, back-emf associate with inner loop electronics, poorly modeled or un-modeled dynamics, and external disturbances (e.g.magnetic, aerodynamic, etc.).As pointing requirements have become more stringent to accomplish missions in space, decoupling dynamic disturbance torques is an attractive solution provided by the physics-based control design methodology.Promising approaches include elimination of virtual-zero references, manipulated input decoupling, which can be augmented with disturbance input decoupling supported by sensor replacement.This chapter introduces these methods of physics-based control.Physics based control is a method that seeks to significantly incorporate the dominant physics of the problem to be controlled into the control design.Some components of the methods include elimination of zero-virtual reference, observers for sensor replacements, manipulated input decoupling, and disturbance-input estimation and decoupling.In addition, it will be shown that cross-axis coupling inherent in the governing dynamics can be eliminated by decoupling a normal part of the physics-based control.Physics-based controls methods produce a idealized feedforward control based on the system dynamics that is easily augmented with adaptive techniques to both improve performance and assist on-orbit system identification. Physics-based controls 2.1 Zero-virtual referencesZero-virtual references are implicit with cascaded control loops.When inner loops reference signals are not designed otherwise, the cascaded topology results in zero-references, where the inner loop states are naturally zero-seeking.It is generally understood that if any control system demands a positive or negative rate, the inner position loop (seeking zero) would essentially be fighting the rate loop, since a positive or negative rate command with quiescent initial conditions dictates non-zero position command.Elimination of the zerovirtual reference may be accomplished by using analytic expressions for both position and rate eliminating the nested, cascaded topology.Using analytic expressions for both position and rate commands implies the utilization of commands that both correspond to achieving www.intechopen.comAdvances in Spacecraft Systems and Orbit Determination 30 the same desired end state, essentially eliminating the conflict between the position and rate commands inherit in the cascaded topology. Manipulated Input Decoupling (MID)Manipulated input is the actual variable that can be modified by a control design.Very often in academic settings control u is the goal of a design, but in reality a voltage command is sent to a control actuator, and this voltage command should be referred to as the true manipulated input.The importance of this distinction lies in the fact that electronics may not properly replicated the desired control u, unless the control designer has accounted for internal disturbance factors like the resistive effects of back-emf (inherit in any electronic device where current is generated and modified in the presence of a magnetic field).The manipulated input signal should be designed to decouple these effects.