This article proposes a deep reinforcement learning method to generate a controller for the STTW robot to complete the ramp jump task. First, the corresponding state space and action space are designed based on the characteristics of ramp jumping. Second, an effective reward function is developed to help train the agent.

Experience the power of the most sophisticated AI technologies unified on a single platform. Empower teams to collaborate, boost productivity, and harness the potential of advanced AI technology. Give it a Try!

Ramp automates tasks, eliminates errors, and uncovers insights with AI designed for finance teams. Meet your partner in finance. Ramp is a natural extension of your team, automating tedious tasks and delivering thoughtful suggestions—all with you in control. Expenses that submit themselves. Reclaim hours lost each month on expense reports.

Drive growth across every team and function with the ramp.ai generative AI platform. Enable seamless alignment with business goals and objectives, identify use cases to maximize value, and bring your transformation to life with a tailored approach for your enterprise.

Abstract: In contrast to attitude control strategies relying solely on the aerial phase, the single-track two-wheeled (STTW) robots can adjust their attitude during both the take-off and the aerial phase in the ramp jump task, enabling a wider range of adjustments to achieve an appropriate landing attitude, which introduces a multi-contact ...

Ramp Co-Founder and CEO Eric Glyman pioneered AI-powered fintech with Paribus in 2015 before building Ramp into a leading finance automation platform. His insights reveal how AI can transform business processes by deeply understanding user needs and automating tedious tasks to enable more strategic work.

2025年3月12日 · AI is Moving from Experimentation to Essential Infrastructure. Ramp’s unique vantage point—processing over $55 billion in corporate spending—provides valuable insights into AI adoption across industries. Glyman noted a dramatic shift in AI usage, from experimental trial-and-error to widespread, integral adoption.

In this work, we propose a framework for learning an end-to-end vision-based whole-body-control parkour policy for humanoid robots that overcomes multiple parkour skills without any motion prior. Using the parkour policy, the humanoid robot can jump on a 0.42m platform, leap over hurdles, 0.8m gaps, and much more.

2021年8月28日 · We present a control method that employs continuous action reinforcement learning techniques for single-track two-wheeled robot control. We design a novel reward function for reinforcement learning, optimize the dimensions of the action space, and enable training under the deep deterministic policy gradient algorithm.

2024年8月1日 · Now, a decade later, Ramp is testing agentic AI to analyze hundreds of thousands of customer calls and generate custom summaries for teams. Team members can ask our AI agent—affectionately named Toby—what our customers have said about any topic and get a detailed digest based on real customer conversations.