Accepted Papers

• Dual-Objective Reinforcement Learning with Novel Hamilton-Jacobi-Bellman Formulations
  William Sharpless, Dylan Hirsch, Sander Tonkens, Nikhil Shinde, and Sylvia Herbert

• Safe Active Navigation and Exploration for Planetary Environments Using Proprioceptive Measurements
  Matthew Jiang, Feifei Qian, and Shipeng Liu

• Nightmare Dreamer: Dreaming About Unsafe States And Planing Ahead
  Oluwatosin Oseni, Shengjie Wang, Jun Zhu, and Micah Corah

• Equitable Optimization Across Preferences: A Maximinimalist Approach to Persistent Monitoring with Preference-based Reinforcement Learning
  Manav Mishra, Peihong Yu, Sujit P.B., and Pratap Tokekar

Contributions

The workshop features two types of calls for contributions: 

1.  call for workshop papers, and

2.  call for provocative propositions.

All contributions should be submitted to rss2025moo(at)gmail.com 

Call for Papers

We invite researchers to contribute short papers (max 4 pages, excl. references) that either are (i) blue sky ideas and position papers that provoke discussion, (ii) new, potentially unpublished results, or (iii) recently published work from robotics venues that may not have been presented at conferences (e.g., journal papers). All papers should be formatted using the RSS conference template.

Topics of interest include, but are not limited to:

• Topics of interest include, but are not limited to:

• Multi-objective optimization and planning in robotics

• Balancing trade-offs in planning, parameter tuning / auto-tuning of controllers and their objectives

• Active perception

• Reward design / designing cost functions

• Multi-objective RL

• Preference learning and personalization of robots

• Inverse reinforcement learning / Inverse optimal control

• Robot design

Papers will undergo single-blind review and thus authors are not required to be anonymous. All accepted papers will be presented at the workshop in (i) a 3 minute pitch talk, and (ii) a 45min poster session. The manuscripts of accepted papers will be made available on this website upon the authors' consent.
Paper review and acceptance will be on a rolling basis, such that early submissions receive an early notification. We hope that this helps to arrange travel plans. However, the final deadline for submission is strict.
Papers should be submitted in PDF format via email to rss2025moo(at)gmail.com

Important Dates

• Paper Submissions Deadline: May 31, 2025 (Anytime on Earth),

• Author Notification: Rolling basis,

• Final version due: June 15, 2025,

• Workshop: June 25, 2025.

Call for Propositions

We invite all workshop participants to provide a proposition in reply to one or several of our three key perspectives and open questions below. The two most thought-provoking propositions will be featured in a 10min "defense" where the proposer can explain their hot-take on the topic and initiate an active debate. The presenters will also be part of the panel discussion that follows directly after.

Key perspectives on current challenges of multi-objective optimization in robotics:

1. Obliviousness: practitioners may simply lack awareness of the multi-objective nature of their problem. E.g., it is easy to follow the (common) process of formulating a cost function as a weighted sum without realizing one is forming a scalarized objective. 

2. Inapplicability: there may be the sense among roboticists that multi-objective optimization does not apply. The multi-objective perspective may appear inapplicable for sequential decision-making problems: when ‘the rubber meets the road’ a robot must autonomously decide (e.g., choose actions) and thus resolve multi-objective ambiguities. 

3. Inaccessibility: tools (both conceptual and computational) are less widely known, so seldom adopted, hence less available. E.g., tools may be specifically targeted for particular problems (e.g., robot design) so,  without aiming to appear narrow, are not adopted more widely.

Open questions:

1. Why don’t we see more people in the robotics community using multi-objective optimization? Can you give an example in your own work of where you only later came to see the true multi-objective view of the problem?

2. What are seldom acknowledged limitations of multi-objective optimization in the robotics literature? What do you consider the implicit “dirty laundry” in papers/work on multi-objective optimization in robotics that you’ve seen?

3.  Will we eventually have the same evolution we see in single objective optimization in the multi-objective case (for instance, there is a lot of focus on certifiable optimization now—are we just a few years behind in the multi-objective case?)

4.  To what extent are multiple objectives equivalent to constraints with tunable bounds? Is there a principled way to decide which undesirable quantities should be treated as constraints or minimized as an objective?

5. If a robot is truly behaving autonomously, it must make a choice when the “rubber meets the road”. When is the right time to resolve the non-uniqueness of solutions to multi-objective optimization problems?

Propositions should be ~200 words in length (per perspective / question). Participants may provide propositions to one or several perspectives and open questions. Propositions can have multiple authors. All submissions will be reviewed by the organizing committee.
Propositions should be submitted in PDF format via email to rss2025moo(at)gmail.com

Important Dates

• Proposition Submissions Deadline: June 24, 2025 (noon local time),

• Authors Notification: June 24, 2025 (end of day),

• Workshop: June 25, 2025.