-
Procedural involute spur gears in Blender 3 with geometry nodes
I love designing functional models for 3D printing with Blender. I know it’s not a common choice and CAD software is usually preferred, and I have tested many but none has offered me the flexibility and speed of Blender for my projects. One issue that I always had was generating gears and bolts. Of course […]
-
Taking advantage of ZFS for smarter Proxmox backups
Let’s say we have a Proxmox cluster running ~30 VMs using ZFS as a storage backend. We want to backup each VM hourly to a remote server, and then replicate these backups to an offsite server. Proxmox Backup Server is nicely integrated into PVE’s web GUI, and can work with ZFS volumes. However, PBS is […]
-
Comparer DPE et consommation a-t-il un sens ?
Récapitulatif des publications à ce sujet : Hello Watt a publié le 4 janvier une étude concluant à un manque de corrélation entre le DPE d’un logement, censé évaluer sa performance énergétique, et sa consommation d’énergie mesurée par les compteurs communicants. Pourquoi cette étude ? Dans sa mission de favoriser la transition énergétique des ménages, […]
-
Proxmox cluster on distant bare metal servers
I run a Proxmox cluster with three nodes that are set up on three rented bare-metal servers from OVH in different datacenters. This is a pretty unusual setup, because bare-metal rental companies to not allow bridging on their network interface. Bridging in a typical racked cluster In this context, bridging means that a single physical […]
-
Predicting Home Energy Consumption, the Data-Driven Way
At Hello Watt, we help residential energy consumers reduce their energy bills through various means. This involves estimating their electricity and gas consumption based on information we collect either over the phone or through a form. We developed Consumption Calculator, a new model that works very similarly to the previous model except its coefficients are optimized automatically based on training data derived from our database.
-
Domain Knowledge Aids in Signal Disaggregation; the Example of the Cumulative Water Heater
In this article we present an unsupervised low-frequency method aimed at detecting and disaggregating the power used by Cumulative Water Heaters (CWH) in residential homes. Our model circumvents the inherent difficulty of unsupervised signal disaggregation by using both the shape of a power spike and its temporal pattern to identify the contribution of CWH reliably. Indeed, many CHWs in France are configured to turn on automatically during off-peak hours only, and we are able to use this domain knowledge to aid peak identification despite the low sampling frequency. In order to test our model, we equipped a home with sensors to record the ground truth consumption of a water heater. We then apply the model to a larger dataset of energy consumption of Hello Watt users consisting of one month of consumption data for 5k homes at 30-minute resolution. In this dataset we successfully identified CWHs in 66.5% of cases where consumers declared using them. Inability of our model to identify CWHs in the consumption signal in the remaining cases is likely due to possible misconfiguration of CWHs, since triggering them during off-peak hours requires specific wiring in the electrical panel of the house. Our model, despite its simplicity, offers promising applications: detection of mis-configured CWHs on off-peak contracts and slow performance degradation.
-
The nightmare of Shingled Magnetic Recording for NAS drives
IT systems are like onions, they have layers. And when the bottom-most layer fails, everything follows. This is the story of our experience with WD Red hard disk drives at Hello Watt, and how even while following good practices of system administration, deceptive advertising can cause your business to lose weeks of work.
-
Optimization of Google Ads bidding taking into account our influence on the market
In this article, we give a brief overview of the Google Ads platform from the point of view of an advertiser as well as its available performance indicators and controls. We show how a simple portfolio optimization method indicates we can increase our ad returns significantly. Finally, we design a more complex model that takes into account market saturation and show that the actual portfolio optimized by Google is very close to the maximum possible return predicted by our model.
-
PBCS: Efficient Exploration and Exploitation Using a Synergy between Reinforcement Learning and Motion Planning
The exploration-exploitation trade-off is at the heart of reinforcement learning (RL). However, most continuous control benchmarks used in recent RL research only require local exploration. This led to the development of algorithms that have basic exploration capabilities, and behave poorly in benchmarks that require more versatile exploration. For instance, as demonstrated in our empirical study, state-of-the-art RL algorithms such as DDPG and TD3 are unable to steer a point mass in even small 2D mazes. In this paper, we propose a new algorithm called ”Plan, Backplay, Chain Skills” (PBCS) that combines motion planning and reinforcement learning to solve hard exploration environments. In a first phase, a motion planning algorithm is used to find a single good trajectory, then an RL algorithm is trained using a curriculum derived from the trajectory, by combining a variant of the Backplay algorithm and skill chaining. We show that this method outperforms state-of-the-art RL algorithms in 2D maze environments of various sizes, and is able to improve on the trajectory obtained by the motion planning phase.
-
The problem with DDPG: understanding failures in deterministic environments with sparse rewards
In environments with continuous state and action spaces, state-of-the-art actor-critic reinforcement learning algorithms can solve very complex problems, yet can also fail in environments that seem trivial, but the reason for such failures is still poorly understood. In this paper, we contribute a formal explanation of these failures in the particular case of sparse reward and deterministic environments. First, using a very elementary control problem, we illustrate that the learning process can get stuck into a fixed point corresponding to a poor solution, especially when the reward is not found very early. Then, generalizing from the studied example, we provide a detailed analysis of the underlying mechanisms which results in a new understanding of one of the convergence regimes of these algorithms.