Deep Reinforcement Learning for Robotic Controls

Digital Twin is enjoying a breakthrough, fueled by advances in the sophisticated use of physics-based simulation models of varying fidelities. Connected and informed by data from the field with the help of artificial intelligence, Leonardo S.p.A. validates the performance of product systems virtually in the early stages of design to accelerate the development of their next generation of products.

The door closing feel is directly experienced by the customer and is typically treated as a subjective functional attribute. Objective targets have also been identified and correlated with subjective feelings, even though the final sign-off happens based on subjective assessment. Door Closing is a complex function involving multiple subsystems, their behavior & interface across parts.

AAM has been providing mechanical drivelines to automotive customers for more than 25 years. As the automobile transitions to electrification, so too must the driveline. AAM has developed a suite of innovative, lightweight, compact and cost-optimized electric drive units and electric beam axles to meet the growing global technology demand as electric vehicles continue to expand over the next decade. Engineering simulation, optimization and validation tools were significant enablers to increase the validation confidence and reduce the time to market for these new products.

The Digital Twin concept is broadly applicable to all Altair customer base wanting to develop better products faster. We will explain how to maximize the value you can get from the Altair Platform for Digital Twins. We will illustrate how this solution helps diverse engineering teams develop better, more complex products faster by avoiding work in silos. This will include proven uses cases for the convergence of simulation and data.

AI-powered real-time melt-pool analytics for accelerated product development and production.

What if you could create 1D simulation models that retain the precision of your 3D analysis while running up to thousands of times faster?

Nowadays, many Machine Learning techniques able to identify dynamic models from data are taking hold. But, can we support the learning phase leveraging our expertise obtaining models with better generalization property? To tackle this need and others we have developed a novel approach, romAI. This application combines AI and system modeling techniques to generate reusable continuous dynamic models extremely time-efficient, hence used as Reduced-Order Models to speed-up system design and optimization or in Digital Twins/Simulators for real-time applications.

The ever-advancing digitization is being driven by the convergence of simulation, data, and HPC. While virtual methods, interoperable tools, and standardized interfaces become state of the art, the engineers and their management will play a significant role to break down silos and benefit from a Digital Thread. In this fireside chat, we will discuss the importance of people to unleash the potential of Digital Twins, how they can combine the different fields of product development and what is still missing to make the connection.

Altair products are quickly evolving into AI-driven products. Starting from the modeling and visualization products, the latest release of Altair HyperWorks includes features like shapeAI, which quickly finds and classifies parts “by shape” inside geometry files or finite element meshes - when you want to locate all of the bolts, shafts, and gears in a complex assembly - by applying machine learning.

Several OEMs currently face the challenge of information silos due to poor integration of models across the product life cycle. Furthermore, there is limited reuse of models between programs and the variation of modeling maturity across various Engineering disciplines results in lack of traceability. In this short presentation, we will look at solutions and workflows that leverages a systems model as a common language of communication while facilitating various types of real time dashboards and visualization to help understand and optimize the entire Mechatronic system performance.

Presented at the ATCx Heavy Equipment in May 2021.

Speaker: Keshav Sundaresh, Global Director – Smart Systems & Mechatronics, Altair

Duration: 20 minutes

For several years now, co-simulation between Multibody Dynamics (MBD) codes like MotionSolve and Discrete Element Method (DEM) software like EDEM has been used for many applications such as vehicles and material transportation. One missing link has been a realistic tire model that not only could interact with the soil material in a realistic way but also show the influence of varying tire pressure, change in contact area and pressure distribution over said contact area.

For very soft material such as deep mud with a high-pressure tire, using a rigid wheel is a decent approximation. But for tires on harder surface, this method has several short comings. For example, a rigid representation of the tire will have neglectable rolling resistance while the pressure distribution on a real tire has its peak well ahead of the centerline of the tire producing a resisting moment to the motion. The contact patch area will in reality not depend on sinkage, but the sinkage is dependent on the contact area that is dependent on load and internal pressure.

This presentation introduces the new PM-FlexTire model that is integrated in EDEM and will work with MotionSolve. Requirements for creating and correlating the tire model is presented together with several application examples on mud, clay and gravel beds.

Presented at the ATCx Heavy Equipment in May 2021.

Speaker: Jesper Slattengren, Technical Fellow, Pratt Miller

Duration: 20 minutes

The chassis of a the RAVO 5 iSeries street sweeper has been a proven design that has remained relatively unchanged since the mid-60’s. However, due to increasing weight and size requirements, as well as possible changes to the driving mechanism, an overhaul of the chassis platform is needed. To develop a new (modular) chassis, which is suited for upcoming and future changes, a CAE-driven design approach is chosen, which is being performed by VIRO. When using computer-aided engineering (CAE), the chief problem is to define or estimate the acting loads and / or boundary conditions on a structure. In case such a structure is dynamically loaded, such a determination is further aggravated. Furthermore, if that structure is a chassis, a wide range of load cases and interconnected systems (e.g. tyres, suspension) effect the multifaceted loading behavior.

By utilizing multibody dynamics (MBD) and creating a MBD-model of the current street sweeper, a first step is taken towards developing a new (modular) chassis. The calculated loads and moments acting on the suspensions and chassis are quantified for e.g. several worst case scenarios, which allows for well-founded decisions in relation to the new design. Furthermore, these load conditions can be used as input for future topographic optimizations and / or stress analyses. An important part, when creating a CAE-model, i.e. in this case a MBD-model, is the validation of that model by experiments and /or analytic formulations. Consequently several calculated load cases have been tested in the field and in mutual cooperation. A subsequent verification shows that the measured accelerations correlate well with the calculated accelerations from the MBD-model.

Presented at the ATCx Heavy Equipment in May 2021.

Speaker: Thijs Romans, Group Leader Engineering Analysis, VIRO

Duration: 20 minutes

G.S. Vidyaprakash presents how Lakshmi Machine Works Ltd. drives the design process with simulation. In his presentation First Time Right Machine Design with CAE Simulation, he discusses reliable simulation techniques to predict & prevent failure modes.

Dr. Fatma Kocer VP, Engineering Data Science at Altair showcases the impact of AI in development environments. In particular on how CAE tools will evolve and Design Exploration is taken to the next level.

Brett Chouinard, President and COO of Altair discusses Development challenges in Industrial machinery and illustrates how Altair's scalable simulation solutions provide the environment for simulation-driven innovation and how machine learning expands the capabilities furter.

Controller strategy expert Lorenzo Moretti presents how to improve Cobot collaboration. In the context of virtual commissioning, he discusses how more meaningful controller design is possible with realistic plant models.

Benoit Pelourdeau presents how Stanley Robotics SAS drives the design process with simulation, to develop the world's first robotic car storage service. Learn from him how the interdisciplinary mechatronic product development team succeeds with accurate virtual prototypes.

Christian Kehrer, Business Development Manager System Modelling, gives a Demo Sessions on Altair Simulation Solutions for faster evaluation of real-world machines, explaining how CNC control optimization is performed to realize toolpath error correction of a CNC milling machine, with system simulation and how more effective controller design with realistic plant models is realized.

Daniel Jauss, Application Engineer CAE, gives a Demo Sessions on Altair Simulation Solutions for faster evaluation of real-world machines, explaining how to improve system understanding with dynamic motion analysis to Identify real behavior of the machine assembly and Identify peak loads.

Industry Innovator Luca Gatti Luca invites you to a virtual coffee break, to see why it is necessary to model and deeply study the physics behind a cup of coffee. Learn why Gruppo Cimbali SpA is creating digital twins of their coffee machines to fully embrace the digital transformation.

Mr. Vijay Zala and Mr. Pragnesh Zala present development methodologies for new economic machine generations and showcase why Jyoti CNC Automation Ltd. sets simulation at the core of the development strategy.

Industry expert Dennis Baum presents how Weber Maschinenbau applies simulation, how simulation can be used along product lifecycle and which benefits arise to customers and internal processes.

Indradev Babu, Managing Director, UCAM PVT LTD, explains how he developed a CNC Jobshop to the largest manufacturer of CNC Rotary tables—presenting different development examples, he explains how Simulation Driven Design helps him to differentiate and what simulation Strategies he implements in the customer-centric development for the following generation Machine Tools at UCAM.

Andrew Dyer, Senior Tech Specialist at Altair in the US, Gabriele Piombo, EngD Partner with Altair at the University of Warwick, and finally Prof. James Marco, Professor of Systems Modelling and Simulation at the University of Warwick, present - Rapid Battery Layout Optimization using 0D / 1D in a Systems Environment. Together they will discuss the use of a 1D Systems approach to simulate a battery module. This includes the cell, cooling plate & busbar and is subjected to a typical extreme duty cycle. A typical event is simulated where the maximum allowable temperature is exceeded which triggers an increase in the cooling flow rate. This system is then optimised.

Dr. Royston Jones, CTO and Dr. Anthony Hahnel, Technical Director, give their keynote presentation during the ATCx Driving Innovation in Electric Powertrain 2021.

In this presentation, MIRDC’s Bionic Intelligent Automatic Guided Vehicle (BI-AGV) is introduced. This “collaborative handling module” has three characteristics of wireless intelligence, flexible use and flexible movement. Through the wireless intelligent collaborative handling system, it can control several automatic guided vehicles (AGVS) in real time, and several vehicles can conduct remote control and serial connection to carry out handling tasks. At the same time, the 360-degree mobile omni-directional wheel design framework is adopted. It has the advantages of flexible use in the area where the traditional unmanned vehicle cannot run smoothly in the indoor narrow space.

Pete Darnell, Vice President, Software Development at Altair discusses why embedded development tools must track the ever-increasing complexity and feature set of microcontrollers as well as demands from new markets like the Cloud based “Internet of Things”. He looks at trends in the microcontroller space and discusses how block diagram model-based development tools can be used to abstract these changes to ease firmware development. He also looks at the thing side of IoT and the challenges of power management to extend battery life and what it will take to provide secure over-the-air firmware updates. The recording is about 19 minutes long, and was presented at the 2020 Global Altair Technology conference.

Altair’s portfolio of simulation-driven design solutions covers – amongst many other disciplines – electromagnetics (EM) and electronic design automation (EDA).

The world is full of bulk and granular materials. From ores being mined, soil being excavated, rocks being conveyed, or powders being processed - over 70% of all industrial processes involve the handling or processing of these challenging materials.

Winning the innovation race for mechatronics products requires a tool set which can support a multi-disciplinary development process from early concept design up to in-service support.

These success stories illustrate how customers are leveraging Altair's Math & Systems technology for Model-Based Development to develop better products, faster. Simulations involve 3D, 1D, and/or 0D modeling approaches based on the integrated use of Altair MotionSolve™, Altair Activate™, and/or Altair Compose™.

Gonçalo Pereira, Principal Applications Engineer at Altair presents at the 2019 UK e-Mobility Seminar. Trade-off Studies between Battery Pack, e-Motor, Range etc. System Model Generation to Explore Sensitivities.

With the prospect of increasing labor shortages among a wide variety of industries, combined with the significant costs of occupational injuries and the ever-increasing pressure to increase productivity, leading robotics company, Sarcos is in a unique position to deploy industrial robotics designed to increase productivity while eliminating injuries, by augmenting rather than replacing human workers. Chris Beaufait, COO of Sarcos, discusses the current robotics landscape, why automation is not the right solution to the problems industries are facing, how Sarcos and its product lineup - including the full body, fully powered Guardian XO exoskeleton - will play a critical role in defining the workforce of tomorrow, and his vision for the robotics industry over the next five to 10 years.

Sendyne, part of the Altair Partner Alliance, provides a virtual, physics-based battery model called CellMod Virtual Battery for real-time co-simulation for individual cells and packs. This presentation provides a brief overview of the battery model and benefits as well as an example of system simulation using Altair Activate using the Functional Mock-up Interface standard for co-simulation.

Presentation by Konstantin Arhiptsov & by Eitan Maler, Simulation Dept. IWI Israel Weapon Industries (IWI) at the ATCx in Israel, Netanya on October 30, 2019. These days, in IWI, complete multi physics simulation is an integrated tool in the development of any new product. The motivation is to completely simulate one or two firing cycles as close to reality. First step is the Multi Body Dynamics simulation to check all mechanism are synchronized and work properly. Second is the explicit simulations - calibrating the mechanical properties of the pistol, in that the springs, contacts, materials and gun powder properties based on one firing cycle. The following is to calibrate the Non Rigid Boundary conditions (NRBC’s). This calibration of boundary conditions which are not completely fixed is crucial to understanding the actual strains and stresses on the parts. One of the approaches was to use known data of the stiffness of arm and wrist, implementing this data into a HyperStudy model to compare and calibrate the results based on a slow motion capturing of a real firing. The results are promising, with high accuracy of the behavior compared to a real capturing of the shooting, up to the point of slider getting to the end of its move – where most of the kinetic energy transform into loads on the frame. The following steps will be to calibrate, using the same method, the return of the slider to it’s original position and perform more than one firing cycle.

Presenter: Ronald Kett, Technical Specialist, Altair

For a Stewart-Gough-Platform (Hexapod), various software tools were used to study and design highly dynamic hydraulic drives together with an overall system control. Calculation of Eigenfrequencies, control design and comparison, hydraulic system design, and overall simulation control were done in Altair Activate, the mechanics of the Stewart-Gough-Platform was taken from a CAD model into Altair Inspire Motion. The co-simulation between control + hydraulics and mechanics was performed using Activate and Altair MotionSolve. Altair HyperView and HyperGraph were used to analyze and visualize the results. With the highly integrated solutions, the results could be achieved within a very short time. The different types of models (linear/simplified/full mechanics/hydraulics) made it possible to start with fast development cycles and finally achieve reliable results.

Presenter: Christian Kehrer, Business Development Manager, Altair

This presentation discusses the multi-disciplinary evaluation of truck platooning, with the lead truck sending out acceleration, braking and steering signals for the following trucks to react accordingly. The benefits address safety requirements, fuel savings, traffic capacity and convenience. The presentation demonstrates why platooning requires a holistic approach in the sense of connecting different modeling and simulation methods for a virtual evaluation of this system of systems.

Presenter: Dario Mangoni, Engineering Professor, University of Parma

In the modern car industry, the advent of hybrid and electric vehicle systems is driving radical changes in the car electronics and software, demanding more and more advanced controlling techniques. Self-stopping, self-starting, ultimately self-driving cars are nowadays possible, because of the multitude of sensors, controller units and actuators making the vehicles “smart”. To simplify and make the interaction between the user and the machine more and more intuitive and user-friendly, a much broader and deeper investigation of different use scenario combined with the human interaction and intervention is critical. In this context, higher-detailed vehicle models are required to provide a valid prototyping tool which can be reliably used to test innovative controlling strategies, such as testing with the Man-In-the-Loop. The Car Real-Time Modelica library proposed here aims at providing a highly valuable tool for the vehicle control system design and test. The key competitive advantages in this approach are in the Maple model-based compiler for supporting high-level of details modeling; the adoption of the Modelica language which allows a transparent and physical approach to the modeling activities and finally the Activate platform which offers real-time capabilities within an environment meant for the signal-based control design. To graphically validate the library results, a visualization framework for realistic real-time simulations that assures high-fidelity scenario in which to test user experience was also realized.

Presenter: Nino Michniok, Mechanical Engineering Student, University of Kaiserslautern

The first part of the presentation shows the detailed process of building the multibody system of an actuated exoskeleton in MotionView/MotionSolve (MV/MS). The required movements are transferred to the corresponding joints by “Motions”. By this the exoskeleton can Stand Up, Walk diagonally across the floor and Sit Down. In the second part the “Motions” in MV/MS are replaced by controllers (position control) whichdeliver a certain torque to actuate the exoskeleton. The main topic here is the implementation of the co-simulation between Activate and MV/MS. In the end the presentation gives a quick outlook of similar works at the University of Applied Sciences Kaiserslautern in Germany.

Presenter: Stefano Benanti, R&D materials engineer, Hutchinson

Battery cooling (BC) systems are frequently composed of several parallel branches, each leading to and away from a series of cooling plates. As a correct flow distribution in each branch and overall pressure drop are a key requirement from every customer, numerical computation is extremely important from the first stages of each project: the number of components and their dimensions have a relevant impact on the total cost and it is thus necessary to quickly provide results already in the Request for Quotation (RFQ) phase.The 3D computation of such cases, albeit feasible, takes a relevant amount of time and makes it more costly (both in terms of computational power and of necessary software licenses) to quickly provide results. The goal is then to develop a quicker method to provide results and allow for the necessary optimization cycles. Altair Activate® was chosen by Hutchinson to develop a library of ROMs representing different circuit components through which is possible to create 1D models able to respond quickly and precisely to such demands.